1 /*************************************************
2 * Perl-Compatible Regular Expressions *
3 *************************************************/
4
5 /* PCRE is a library of functions to support regular expressions whose syntax
6 and semantics are as close as possible to those of the Perl 5 language.
7
8 Written by Philip Hazel
9 Copyright (c) 1997-2021 University of Cambridge
10
11 -----------------------------------------------------------------------------
12 Redistribution and use in source and binary forms, with or without
13 modification, are permitted provided that the following conditions are met:
14
15 * Redistributions of source code must retain the above copyright notice,
16 this list of conditions and the following disclaimer.
17
18 * Redistributions in binary form must reproduce the above copyright
19 notice, this list of conditions and the following disclaimer in the
20 documentation and/or other materials provided with the distribution.
21
22 * Neither the name of the University of Cambridge nor the names of its
23 contributors may be used to endorse or promote products derived from
24 this software without specific prior written permission.
25
26 THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS"
27 AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
28 IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
29 ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE
30 LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
31 CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
32 SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS
33 INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN
34 CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
35 ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
36 POSSIBILITY OF SUCH DAMAGE.
37 -----------------------------------------------------------------------------
38 */
39
40
41 /* This module contains the external function pcre_compile(), along with
42 supporting internal functions that are not used by other modules. */
43
44
45 #ifdef HAVE_CONFIG_H
46 #include "config.h"
47 #endif
48
49 #define NLBLOCK cd /* Block containing newline information */
50 #define PSSTART start_pattern /* Field containing pattern start */
51 #define PSEND end_pattern /* Field containing pattern end */
52
53 #include "pcre_internal.h"
54
55
56 /* When PCRE_DEBUG is defined, we need the pcre(16|32)_printint() function, which
57 is also used by pcretest. PCRE_DEBUG is not defined when building a production
58 library. We do not need to select pcre16_printint.c specially, because the
59 COMPILE_PCREx macro will already be appropriately set. */
60
61 #ifdef PCRE_DEBUG
62 /* pcre_printint.c should not include any headers */
63 #define PCRE_INCLUDED
64 #include "pcre_printint.c"
65 #undef PCRE_INCLUDED
66 #endif
67
68
69 /* Macro for setting individual bits in class bitmaps. */
70
71 #define SETBIT(a,b) a[(b)/8] |= (1U << ((b)&7))
72
73 /* Maximum length value to check against when making sure that the integer that
74 holds the compiled pattern length does not overflow. We make it a bit less than
75 INT_MAX to allow for adding in group terminating bytes, so that we don't have
76 to check them every time. */
77
78 #define OFLOW_MAX (INT_MAX - 20)
79
80 /* Definitions to allow mutual recursion */
81
82 static int
83 add_list_to_class(pcre_uint8 *, pcre_uchar **, int, compile_data *,
84 const pcre_uint32 *, unsigned int);
85
86 static BOOL
87 compile_regex(int, pcre_uchar **, const pcre_uchar **, int *, BOOL, BOOL, int, int,
88 pcre_uint32 *, pcre_int32 *, pcre_uint32 *, pcre_int32 *, branch_chain *,
89 compile_data *, int *);
90
91
92
93 /*************************************************
94 * Code parameters and static tables *
95 *************************************************/
96
97 /* This value specifies the size of stack workspace that is used during the
98 first pre-compile phase that determines how much memory is required. The regex
99 is partly compiled into this space, but the compiled parts are discarded as
100 soon as they can be, so that hopefully there will never be an overrun. The code
101 does, however, check for an overrun. The largest amount I've seen used is 218,
102 so this number is very generous.
103
104 The same workspace is used during the second, actual compile phase for
105 remembering forward references to groups so that they can be filled in at the
106 end. Each entry in this list occupies LINK_SIZE bytes, so even when LINK_SIZE
107 is 4 there is plenty of room for most patterns. However, the memory can get
108 filled up by repetitions of forward references, for example patterns like
109 /(?1){0,1999}(b)/, and one user did hit the limit. The code has been changed so
110 that the workspace is expanded using malloc() in this situation. The value
111 below is therefore a minimum, and we put a maximum on it for safety. The
112 minimum is now also defined in terms of LINK_SIZE so that the use of malloc()
113 kicks in at the same number of forward references in all cases. */
114
115 #define COMPILE_WORK_SIZE (2048*LINK_SIZE)
116 #define COMPILE_WORK_SIZE_MAX (100*COMPILE_WORK_SIZE)
117
118 /* This value determines the size of the initial vector that is used for
119 remembering named groups during the pre-compile. It is allocated on the stack,
120 but if it is too small, it is expanded using malloc(), in a similar way to the
121 workspace. The value is the number of slots in the list. */
122
123 #define NAMED_GROUP_LIST_SIZE 20
124
125 /* The overrun tests check for a slightly smaller size so that they detect the
126 overrun before it actually does run off the end of the data block. */
127
128 #define WORK_SIZE_SAFETY_MARGIN (100)
129
130 /* Private flags added to firstchar and reqchar. */
131
132 #define REQ_CASELESS (1U << 0) /* Indicates caselessness */
133 #define REQ_VARY (1U << 1) /* Reqchar followed non-literal item */
134 /* Negative values for the firstchar and reqchar flags */
135 #define REQ_UNSET (-2)
136 #define REQ_NONE (-1)
137
138 /* Repeated character flags. */
139
140 #define UTF_LENGTH 0x10000000l /* The char contains its length. */
141
142 /* Table for handling escaped characters in the range '0'-'z'. Positive returns
143 are simple data values; negative values are for special things like \d and so
144 on. Zero means further processing is needed (for things like \x), or the escape
145 is invalid. */
146
147 #ifndef EBCDIC
148
149 /* This is the "normal" table for ASCII systems or for EBCDIC systems running
150 in UTF-8 mode. */
151
152 static const short int escapes[] = {
153 0, 0,
154 0, 0,
155 0, 0,
156 0, 0,
157 0, 0,
158 CHAR_COLON, CHAR_SEMICOLON,
159 CHAR_LESS_THAN_SIGN, CHAR_EQUALS_SIGN,
160 CHAR_GREATER_THAN_SIGN, CHAR_QUESTION_MARK,
161 CHAR_COMMERCIAL_AT, -ESC_A,
162 -ESC_B, -ESC_C,
163 -ESC_D, -ESC_E,
164 0, -ESC_G,
165 -ESC_H, 0,
166 0, -ESC_K,
167 0, 0,
168 -ESC_N, 0,
169 -ESC_P, -ESC_Q,
170 -ESC_R, -ESC_S,
171 0, 0,
172 -ESC_V, -ESC_W,
173 -ESC_X, 0,
174 -ESC_Z, CHAR_LEFT_SQUARE_BRACKET,
175 CHAR_BACKSLASH, CHAR_RIGHT_SQUARE_BRACKET,
176 CHAR_CIRCUMFLEX_ACCENT, CHAR_UNDERSCORE,
177 CHAR_GRAVE_ACCENT, ESC_a,
178 -ESC_b, 0,
179 -ESC_d, ESC_e,
180 ESC_f, 0,
181 -ESC_h, 0,
182 0, -ESC_k,
183 0, 0,
184 ESC_n, 0,
185 -ESC_p, 0,
186 ESC_r, -ESC_s,
187 ESC_tee, 0,
188 -ESC_v, -ESC_w,
189 0, 0,
190 -ESC_z
191 };
192
193 #else
194
195 /* This is the "abnormal" table for EBCDIC systems without UTF-8 support. */
196
197 static const short int escapes[] = {
198 /* 48 */ 0, 0, 0, '.', '<', '(', '+', '|',
199 /* 50 */ '&', 0, 0, 0, 0, 0, 0, 0,
200 /* 58 */ 0, 0, '!', '$', '*', ')', ';', '~',
201 /* 60 */ '-', '/', 0, 0, 0, 0, 0, 0,
202 /* 68 */ 0, 0, '|', ',', '%', '_', '>', '?',
203 /* 70 */ 0, 0, 0, 0, 0, 0, 0, 0,
204 /* 78 */ 0, '`', ':', '#', '@', '\'', '=', '"',
205 /* 80 */ 0, ESC_a, -ESC_b, 0, -ESC_d, ESC_e, ESC_f, 0,
206 /* 88 */-ESC_h, 0, 0, '{', 0, 0, 0, 0,
207 /* 90 */ 0, 0, -ESC_k, 0, 0, ESC_n, 0, -ESC_p,
208 /* 98 */ 0, ESC_r, 0, '}', 0, 0, 0, 0,
209 /* A0 */ 0, '~', -ESC_s, ESC_tee, 0,-ESC_v, -ESC_w, 0,
210 /* A8 */ 0,-ESC_z, 0, 0, 0, '[', 0, 0,
211 /* B0 */ 0, 0, 0, 0, 0, 0, 0, 0,
212 /* B8 */ 0, 0, 0, 0, 0, ']', '=', '-',
213 /* C0 */ '{',-ESC_A, -ESC_B, -ESC_C, -ESC_D,-ESC_E, 0, -ESC_G,
214 /* C8 */-ESC_H, 0, 0, 0, 0, 0, 0, 0,
215 /* D0 */ '}', 0, -ESC_K, 0, 0,-ESC_N, 0, -ESC_P,
216 /* D8 */-ESC_Q,-ESC_R, 0, 0, 0, 0, 0, 0,
217 /* E0 */ '\\', 0, -ESC_S, 0, 0,-ESC_V, -ESC_W, -ESC_X,
218 /* E8 */ 0,-ESC_Z, 0, 0, 0, 0, 0, 0,
219 /* F0 */ 0, 0, 0, 0, 0, 0, 0, 0,
220 /* F8 */ 0, 0, 0, 0, 0, 0, 0, 0
221 };
222
223 /* We also need a table of characters that may follow \c in an EBCDIC
224 environment for characters 0-31. */
225
226 static unsigned char ebcdic_escape_c[] = "@ABCDEFGHIJKLMNOPQRSTUVWXYZ[\\]^_";
227
228 #endif
229
230
231 /* Table of special "verbs" like (*PRUNE). This is a short table, so it is
232 searched linearly. Put all the names into a single string, in order to reduce
233 the number of relocations when a shared library is dynamically linked. The
234 string is built from string macros so that it works in UTF-8 mode on EBCDIC
235 platforms. */
236
237 typedef struct verbitem {
238 int len; /* Length of verb name */
239 int op; /* Op when no arg, or -1 if arg mandatory */
240 int op_arg; /* Op when arg present, or -1 if not allowed */
241 } verbitem;
242
243 static const char verbnames[] =
244 "\0" /* Empty name is a shorthand for MARK */
245 STRING_MARK0
246 STRING_ACCEPT0
247 STRING_COMMIT0
248 STRING_F0
249 STRING_FAIL0
250 STRING_PRUNE0
251 STRING_SKIP0
252 STRING_THEN;
253
254 static const verbitem verbs[] = {
255 { 0, -1, OP_MARK },
256 { 4, -1, OP_MARK },
257 { 6, OP_ACCEPT, -1 },
258 { 6, OP_COMMIT, -1 },
259 { 1, OP_FAIL, -1 },
260 { 4, OP_FAIL, -1 },
261 { 5, OP_PRUNE, OP_PRUNE_ARG },
262 { 4, OP_SKIP, OP_SKIP_ARG },
263 { 4, OP_THEN, OP_THEN_ARG }
264 };
265
266 static const int verbcount = sizeof(verbs)/sizeof(verbitem);
267
268
269 /* Substitutes for [[:<:]] and [[:>:]], which mean start and end of word in
270 another regex library. */
271
272 static const pcre_uchar sub_start_of_word[] = {
273 CHAR_BACKSLASH, CHAR_b, CHAR_LEFT_PARENTHESIS, CHAR_QUESTION_MARK,
274 CHAR_EQUALS_SIGN, CHAR_BACKSLASH, CHAR_w, CHAR_RIGHT_PARENTHESIS, '\0' };
275
276 static const pcre_uchar sub_end_of_word[] = {
277 CHAR_BACKSLASH, CHAR_b, CHAR_LEFT_PARENTHESIS, CHAR_QUESTION_MARK,
278 CHAR_LESS_THAN_SIGN, CHAR_EQUALS_SIGN, CHAR_BACKSLASH, CHAR_w,
279 CHAR_RIGHT_PARENTHESIS, '\0' };
280
281
282 /* Tables of names of POSIX character classes and their lengths. The names are
283 now all in a single string, to reduce the number of relocations when a shared
284 library is dynamically loaded. The list of lengths is terminated by a zero
285 length entry. The first three must be alpha, lower, upper, as this is assumed
286 for handling case independence. The indices for graph, print, and punct are
287 needed, so identify them. */
288
289 static const char posix_names[] =
290 STRING_alpha0 STRING_lower0 STRING_upper0 STRING_alnum0
291 STRING_ascii0 STRING_blank0 STRING_cntrl0 STRING_digit0
292 STRING_graph0 STRING_print0 STRING_punct0 STRING_space0
293 STRING_word0 STRING_xdigit;
294
295 static const pcre_uint8 posix_name_lengths[] = {
296 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 4, 6, 0 };
297
298 #define PC_GRAPH 8
299 #define PC_PRINT 9
300 #define PC_PUNCT 10
301
302
303 /* Table of class bit maps for each POSIX class. Each class is formed from a
304 base map, with an optional addition or removal of another map. Then, for some
305 classes, there is some additional tweaking: for [:blank:] the vertical space
306 characters are removed, and for [:alpha:] and [:alnum:] the underscore
307 character is removed. The triples in the table consist of the base map offset,
308 second map offset or -1 if no second map, and a non-negative value for map
309 addition or a negative value for map subtraction (if there are two maps). The
310 absolute value of the third field has these meanings: 0 => no tweaking, 1 =>
311 remove vertical space characters, 2 => remove underscore. */
312
313 static const int posix_class_maps[] = {
314 cbit_word, cbit_digit, -2, /* alpha */
315 cbit_lower, -1, 0, /* lower */
316 cbit_upper, -1, 0, /* upper */
317 cbit_word, -1, 2, /* alnum - word without underscore */
318 cbit_print, cbit_cntrl, 0, /* ascii */
319 cbit_space, -1, 1, /* blank - a GNU extension */
320 cbit_cntrl, -1, 0, /* cntrl */
321 cbit_digit, -1, 0, /* digit */
322 cbit_graph, -1, 0, /* graph */
323 cbit_print, -1, 0, /* print */
324 cbit_punct, -1, 0, /* punct */
325 cbit_space, -1, 0, /* space */
326 cbit_word, -1, 0, /* word - a Perl extension */
327 cbit_xdigit,-1, 0 /* xdigit */
328 };
329
330 /* Table of substitutes for \d etc when PCRE_UCP is set. They are replaced by
331 Unicode property escapes. */
332
333 #ifdef SUPPORT_UCP
334 static const pcre_uchar string_PNd[] = {
335 CHAR_BACKSLASH, CHAR_P, CHAR_LEFT_CURLY_BRACKET,
336 CHAR_N, CHAR_d, CHAR_RIGHT_CURLY_BRACKET, '\0' };
337 static const pcre_uchar string_pNd[] = {
338 CHAR_BACKSLASH, CHAR_p, CHAR_LEFT_CURLY_BRACKET,
339 CHAR_N, CHAR_d, CHAR_RIGHT_CURLY_BRACKET, '\0' };
340 static const pcre_uchar string_PXsp[] = {
341 CHAR_BACKSLASH, CHAR_P, CHAR_LEFT_CURLY_BRACKET,
342 CHAR_X, CHAR_s, CHAR_p, CHAR_RIGHT_CURLY_BRACKET, '\0' };
343 static const pcre_uchar string_pXsp[] = {
344 CHAR_BACKSLASH, CHAR_p, CHAR_LEFT_CURLY_BRACKET,
345 CHAR_X, CHAR_s, CHAR_p, CHAR_RIGHT_CURLY_BRACKET, '\0' };
346 static const pcre_uchar string_PXwd[] = {
347 CHAR_BACKSLASH, CHAR_P, CHAR_LEFT_CURLY_BRACKET,
348 CHAR_X, CHAR_w, CHAR_d, CHAR_RIGHT_CURLY_BRACKET, '\0' };
349 static const pcre_uchar string_pXwd[] = {
350 CHAR_BACKSLASH, CHAR_p, CHAR_LEFT_CURLY_BRACKET,
351 CHAR_X, CHAR_w, CHAR_d, CHAR_RIGHT_CURLY_BRACKET, '\0' };
352
353 static const pcre_uchar *substitutes[] = {
354 string_PNd, /* \D */
355 string_pNd, /* \d */
356 string_PXsp, /* \S */ /* Xsp is Perl space, but from 8.34, Perl */
357 string_pXsp, /* \s */ /* space and POSIX space are the same. */
358 string_PXwd, /* \W */
359 string_pXwd /* \w */
360 };
361
362 /* The POSIX class substitutes must be in the order of the POSIX class names,
363 defined above, and there are both positive and negative cases. NULL means no
364 general substitute of a Unicode property escape (\p or \P). However, for some
365 POSIX classes (e.g. graph, print, punct) a special property code is compiled
366 directly. */
367
368 static const pcre_uchar string_pL[] = {
369 CHAR_BACKSLASH, CHAR_p, CHAR_LEFT_CURLY_BRACKET,
370 CHAR_L, CHAR_RIGHT_CURLY_BRACKET, '\0' };
371 static const pcre_uchar string_pLl[] = {
372 CHAR_BACKSLASH, CHAR_p, CHAR_LEFT_CURLY_BRACKET,
373 CHAR_L, CHAR_l, CHAR_RIGHT_CURLY_BRACKET, '\0' };
374 static const pcre_uchar string_pLu[] = {
375 CHAR_BACKSLASH, CHAR_p, CHAR_LEFT_CURLY_BRACKET,
376 CHAR_L, CHAR_u, CHAR_RIGHT_CURLY_BRACKET, '\0' };
377 static const pcre_uchar string_pXan[] = {
378 CHAR_BACKSLASH, CHAR_p, CHAR_LEFT_CURLY_BRACKET,
379 CHAR_X, CHAR_a, CHAR_n, CHAR_RIGHT_CURLY_BRACKET, '\0' };
380 static const pcre_uchar string_h[] = {
381 CHAR_BACKSLASH, CHAR_h, '\0' };
382 static const pcre_uchar string_pXps[] = {
383 CHAR_BACKSLASH, CHAR_p, CHAR_LEFT_CURLY_BRACKET,
384 CHAR_X, CHAR_p, CHAR_s, CHAR_RIGHT_CURLY_BRACKET, '\0' };
385 static const pcre_uchar string_PL[] = {
386 CHAR_BACKSLASH, CHAR_P, CHAR_LEFT_CURLY_BRACKET,
387 CHAR_L, CHAR_RIGHT_CURLY_BRACKET, '\0' };
388 static const pcre_uchar string_PLl[] = {
389 CHAR_BACKSLASH, CHAR_P, CHAR_LEFT_CURLY_BRACKET,
390 CHAR_L, CHAR_l, CHAR_RIGHT_CURLY_BRACKET, '\0' };
391 static const pcre_uchar string_PLu[] = {
392 CHAR_BACKSLASH, CHAR_P, CHAR_LEFT_CURLY_BRACKET,
393 CHAR_L, CHAR_u, CHAR_RIGHT_CURLY_BRACKET, '\0' };
394 static const pcre_uchar string_PXan[] = {
395 CHAR_BACKSLASH, CHAR_P, CHAR_LEFT_CURLY_BRACKET,
396 CHAR_X, CHAR_a, CHAR_n, CHAR_RIGHT_CURLY_BRACKET, '\0' };
397 static const pcre_uchar string_H[] = {
398 CHAR_BACKSLASH, CHAR_H, '\0' };
399 static const pcre_uchar string_PXps[] = {
400 CHAR_BACKSLASH, CHAR_P, CHAR_LEFT_CURLY_BRACKET,
401 CHAR_X, CHAR_p, CHAR_s, CHAR_RIGHT_CURLY_BRACKET, '\0' };
402
403 static const pcre_uchar *posix_substitutes[] = {
404 string_pL, /* alpha */
405 string_pLl, /* lower */
406 string_pLu, /* upper */
407 string_pXan, /* alnum */
408 NULL, /* ascii */
409 string_h, /* blank */
410 NULL, /* cntrl */
411 string_pNd, /* digit */
412 NULL, /* graph */
413 NULL, /* print */
414 NULL, /* punct */
415 string_pXps, /* space */ /* Xps is POSIX space, but from 8.34 */
416 string_pXwd, /* word */ /* Perl and POSIX space are the same */
417 NULL, /* xdigit */
418 /* Negated cases */
419 string_PL, /* ^alpha */
420 string_PLl, /* ^lower */
421 string_PLu, /* ^upper */
422 string_PXan, /* ^alnum */
423 NULL, /* ^ascii */
424 string_H, /* ^blank */
425 NULL, /* ^cntrl */
426 string_PNd, /* ^digit */
427 NULL, /* ^graph */
428 NULL, /* ^print */
429 NULL, /* ^punct */
430 string_PXps, /* ^space */ /* Xps is POSIX space, but from 8.34 */
431 string_PXwd, /* ^word */ /* Perl and POSIX space are the same */
432 NULL /* ^xdigit */
433 };
434 #define POSIX_SUBSIZE (sizeof(posix_substitutes) / sizeof(pcre_uchar *))
435 #endif
436
437 #define STRING(a) # a
438 #define XSTRING(s) STRING(s)
439
440 /* The texts of compile-time error messages. These are "char *" because they
441 are passed to the outside world. Do not ever re-use any error number, because
442 they are documented. Always add a new error instead. Messages marked DEAD below
443 are no longer used. This used to be a table of strings, but in order to reduce
444 the number of relocations needed when a shared library is loaded dynamically,
445 it is now one long string. We cannot use a table of offsets, because the
446 lengths of inserts such as XSTRING(MAX_NAME_SIZE) are not known. Instead, we
447 simply count through to the one we want - this isn't a performance issue
448 because these strings are used only when there is a compilation error.
449
450 Each substring ends with \0 to insert a null character. This includes the final
451 substring, so that the whole string ends with \0\0, which can be detected when
452 counting through. */
453
454 static const char error_texts[] =
455 "no error\0"
456 "\\ at end of pattern\0"
457 "\\c at end of pattern\0"
458 "unrecognized character follows \\\0"
459 "numbers out of order in {} quantifier\0"
460 /* 5 */
461 "number too big in {} quantifier\0"
462 "missing terminating ] for character class\0"
463 "invalid escape sequence in character class\0"
464 "range out of order in character class\0"
465 "nothing to repeat\0"
466 /* 10 */
467 "internal error: invalid forward reference offset\0"
468 "internal error: unexpected repeat\0"
469 "unrecognized character after (? or (?-\0"
470 "POSIX named classes are supported only within a class\0"
471 "missing )\0"
472 /* 15 */
473 "reference to non-existent subpattern\0"
474 "erroffset passed as NULL\0"
475 "unknown option bit(s) set\0"
476 "missing ) after comment\0"
477 "parentheses nested too deeply\0" /** DEAD **/
478 /* 20 */
479 "regular expression is too large\0"
480 "failed to get memory\0"
481 "unmatched parentheses\0"
482 "internal error: code overflow\0"
483 "unrecognized character after (?<\0"
484 /* 25 */
485 "lookbehind assertion is not fixed length\0"
486 "malformed number or name after (?(\0"
487 "conditional group contains more than two branches\0"
488 "assertion expected after (?( or (?(?C)\0"
489 "(?R or (?[+-]digits must be followed by )\0"
490 /* 30 */
491 "unknown POSIX class name\0"
492 "POSIX collating elements are not supported\0"
493 "this version of PCRE is compiled without UTF support\0"
494 "spare error\0" /** DEAD **/
495 "character value in \\x{} or \\o{} is too large\0"
496 /* 35 */
497 "invalid condition (?(0)\0"
498 "\\C not allowed in lookbehind assertion\0"
499 "PCRE does not support \\L, \\l, \\N{name}, \\U, or \\u\0"
500 "number after (?C is > 255\0"
501 "closing ) for (?C expected\0"
502 /* 40 */
503 "recursive call could loop indefinitely\0"
504 "unrecognized character after (?P\0"
505 "syntax error in subpattern name (missing terminator)\0"
506 "two named subpatterns have the same name\0"
507 "invalid UTF-8 string\0"
508 /* 45 */
509 "support for \\P, \\p, and \\X has not been compiled\0"
510 "malformed \\P or \\p sequence\0"
511 "unknown property name after \\P or \\p\0"
512 "subpattern name is too long (maximum " XSTRING(MAX_NAME_SIZE) " characters)\0"
513 "too many named subpatterns (maximum " XSTRING(MAX_NAME_COUNT) ")\0"
514 /* 50 */
515 "repeated subpattern is too long\0" /** DEAD **/
516 "octal value is greater than \\377 in 8-bit non-UTF-8 mode\0"
517 "internal error: overran compiling workspace\0"
518 "internal error: previously-checked referenced subpattern not found\0"
519 "DEFINE group contains more than one branch\0"
520 /* 55 */
521 "repeating a DEFINE group is not allowed\0" /** DEAD **/
522 "inconsistent NEWLINE options\0"
523 "\\g is not followed by a braced, angle-bracketed, or quoted name/number or by a plain number\0"
524 "a numbered reference must not be zero\0"
525 "an argument is not allowed for (*ACCEPT), (*FAIL), or (*COMMIT)\0"
526 /* 60 */
527 "(*VERB) not recognized or malformed\0"
528 "number is too big\0"
529 "subpattern name expected\0"
530 "digit expected after (?+\0"
531 "] is an invalid data character in JavaScript compatibility mode\0"
532 /* 65 */
533 "different names for subpatterns of the same number are not allowed\0"
534 "(*MARK) must have an argument\0"
535 "this version of PCRE is not compiled with Unicode property support\0"
536 #ifndef EBCDIC
537 "\\c must be followed by an ASCII character\0"
538 #else
539 "\\c must be followed by a letter or one of [\\]^_?\0"
540 #endif
541 "\\k is not followed by a braced, angle-bracketed, or quoted name\0"
542 /* 70 */
543 "internal error: unknown opcode in find_fixedlength()\0"
544 "\\N is not supported in a class\0"
545 "too many forward references\0"
546 "disallowed Unicode code point (>= 0xd800 && <= 0xdfff)\0"
547 "invalid UTF-16 string\0"
548 /* 75 */
549 "name is too long in (*MARK), (*PRUNE), (*SKIP), or (*THEN)\0"
550 "character value in \\u.... sequence is too large\0"
551 "invalid UTF-32 string\0"
552 "setting UTF is disabled by the application\0"
553 "non-hex character in \\x{} (closing brace missing?)\0"
554 /* 80 */
555 "non-octal character in \\o{} (closing brace missing?)\0"
556 "missing opening brace after \\o\0"
557 "parentheses are too deeply nested\0"
558 "invalid range in character class\0"
559 "group name must start with a non-digit\0"
560 /* 85 */
561 "parentheses are too deeply nested (stack check)\0"
562 "digits missing in \\x{} or \\o{}\0"
563 "regular expression is too complicated\0"
564 ;
565
566 /* Table to identify digits and hex digits. This is used when compiling
567 patterns. Note that the tables in chartables are dependent on the locale, and
568 may mark arbitrary characters as digits - but the PCRE compiling code expects
569 to handle only 0-9, a-z, and A-Z as digits when compiling. That is why we have
570 a private table here. It costs 256 bytes, but it is a lot faster than doing
571 character value tests (at least in some simple cases I timed), and in some
572 applications one wants PCRE to compile efficiently as well as match
573 efficiently.
574
575 For convenience, we use the same bit definitions as in chartables:
576
577 0x04 decimal digit
578 0x08 hexadecimal digit
579
580 Then we can use ctype_digit and ctype_xdigit in the code. */
581
582 /* Using a simple comparison for decimal numbers rather than a memory read
583 is much faster, and the resulting code is simpler (the compiler turns it
584 into a subtraction and unsigned comparison). */
585
586 #define IS_DIGIT(x) ((x) >= CHAR_0 && (x) <= CHAR_9)
587
588 #ifndef EBCDIC
589
590 /* This is the "normal" case, for ASCII systems, and EBCDIC systems running in
591 UTF-8 mode. */
592
593 static const pcre_uint8 digitab[] =
594 {
595 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 0- 7 */
596 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 8- 15 */
597 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 16- 23 */
598 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 24- 31 */
599 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* - ' */
600 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* ( - / */
601 0x0c,0x0c,0x0c,0x0c,0x0c,0x0c,0x0c,0x0c, /* 0 - 7 */
602 0x0c,0x0c,0x00,0x00,0x00,0x00,0x00,0x00, /* 8 - ? */
603 0x00,0x08,0x08,0x08,0x08,0x08,0x08,0x00, /* @ - G */
604 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* H - O */
605 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* P - W */
606 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* X - _ */
607 0x00,0x08,0x08,0x08,0x08,0x08,0x08,0x00, /* ` - g */
608 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* h - o */
609 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* p - w */
610 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* x -127 */
611 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 128-135 */
612 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 136-143 */
613 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 144-151 */
614 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 152-159 */
615 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 160-167 */
616 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 168-175 */
617 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 176-183 */
618 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 184-191 */
619 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 192-199 */
620 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 200-207 */
621 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 208-215 */
622 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 216-223 */
623 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 224-231 */
624 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 232-239 */
625 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 240-247 */
626 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00};/* 248-255 */
627
628 #else
629
630 /* This is the "abnormal" case, for EBCDIC systems not running in UTF-8 mode. */
631
632 static const pcre_uint8 digitab[] =
633 {
634 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 0- 7 0 */
635 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 8- 15 */
636 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 16- 23 10 */
637 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 24- 31 */
638 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 32- 39 20 */
639 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 40- 47 */
640 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 48- 55 30 */
641 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 56- 63 */
642 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* - 71 40 */
643 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 72- | */
644 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* & - 87 50 */
645 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 88- 95 */
646 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* - -103 60 */
647 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 104- ? */
648 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 112-119 70 */
649 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 120- " */
650 0x00,0x08,0x08,0x08,0x08,0x08,0x08,0x00, /* 128- g 80 */
651 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* h -143 */
652 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 144- p 90 */
653 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* q -159 */
654 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 160- x A0 */
655 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* y -175 */
656 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* ^ -183 B0 */
657 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 184-191 */
658 0x00,0x08,0x08,0x08,0x08,0x08,0x08,0x00, /* { - G C0 */
659 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* H -207 */
660 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* } - P D0 */
661 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* Q -223 */
662 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* \ - X E0 */
663 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* Y -239 */
664 0x0c,0x0c,0x0c,0x0c,0x0c,0x0c,0x0c,0x0c, /* 0 - 7 F0 */
665 0x0c,0x0c,0x00,0x00,0x00,0x00,0x00,0x00};/* 8 -255 */
666
667 static const pcre_uint8 ebcdic_chartab[] = { /* chartable partial dup */
668 0x80,0x00,0x00,0x00,0x00,0x01,0x00,0x00, /* 0- 7 */
669 0x00,0x00,0x00,0x00,0x01,0x01,0x00,0x00, /* 8- 15 */
670 0x00,0x00,0x00,0x00,0x00,0x01,0x00,0x00, /* 16- 23 */
671 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 24- 31 */
672 0x00,0x00,0x00,0x00,0x00,0x01,0x00,0x00, /* 32- 39 */
673 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 40- 47 */
674 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 48- 55 */
675 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 56- 63 */
676 0x01,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* - 71 */
677 0x00,0x00,0x00,0x80,0x00,0x80,0x80,0x80, /* 72- | */
678 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* & - 87 */
679 0x00,0x00,0x00,0x80,0x80,0x80,0x00,0x00, /* 88- 95 */
680 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* - -103 */
681 0x00,0x00,0x00,0x00,0x00,0x10,0x00,0x80, /* 104- ? */
682 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 112-119 */
683 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 120- " */
684 0x00,0x1a,0x1a,0x1a,0x1a,0x1a,0x1a,0x12, /* 128- g */
685 0x12,0x12,0x00,0x00,0x00,0x00,0x00,0x00, /* h -143 */
686 0x00,0x12,0x12,0x12,0x12,0x12,0x12,0x12, /* 144- p */
687 0x12,0x12,0x00,0x00,0x00,0x00,0x00,0x00, /* q -159 */
688 0x00,0x00,0x12,0x12,0x12,0x12,0x12,0x12, /* 160- x */
689 0x12,0x12,0x00,0x00,0x00,0x00,0x00,0x00, /* y -175 */
690 0x80,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* ^ -183 */
691 0x00,0x00,0x80,0x00,0x00,0x00,0x00,0x00, /* 184-191 */
692 0x80,0x1a,0x1a,0x1a,0x1a,0x1a,0x1a,0x12, /* { - G */
693 0x12,0x12,0x00,0x00,0x00,0x00,0x00,0x00, /* H -207 */
694 0x00,0x12,0x12,0x12,0x12,0x12,0x12,0x12, /* } - P */
695 0x12,0x12,0x00,0x00,0x00,0x00,0x00,0x00, /* Q -223 */
696 0x00,0x00,0x12,0x12,0x12,0x12,0x12,0x12, /* \ - X */
697 0x12,0x12,0x00,0x00,0x00,0x00,0x00,0x00, /* Y -239 */
698 0x1c,0x1c,0x1c,0x1c,0x1c,0x1c,0x1c,0x1c, /* 0 - 7 */
699 0x1c,0x1c,0x00,0x00,0x00,0x00,0x00,0x00};/* 8 -255 */
700 #endif
701
702
703 /* This table is used to check whether auto-possessification is possible
704 between adjacent character-type opcodes. The left-hand (repeated) opcode is
705 used to select the row, and the right-hand opcode is use to select the column.
706 A value of 1 means that auto-possessification is OK. For example, the second
707 value in the first row means that \D+\d can be turned into \D++\d.
708
709 The Unicode property types (\P and \p) have to be present to fill out the table
710 because of what their opcode values are, but the table values should always be
711 zero because property types are handled separately in the code. The last four
712 columns apply to items that cannot be repeated, so there is no need to have
713 rows for them. Note that OP_DIGIT etc. are generated only when PCRE_UCP is
714 *not* set. When it is set, \d etc. are converted into OP_(NOT_)PROP codes. */
715
716 #define APTROWS (LAST_AUTOTAB_LEFT_OP - FIRST_AUTOTAB_OP + 1)
717 #define APTCOLS (LAST_AUTOTAB_RIGHT_OP - FIRST_AUTOTAB_OP + 1)
718
719 static const pcre_uint8 autoposstab[APTROWS][APTCOLS] = {
720 /* \D \d \S \s \W \w . .+ \C \P \p \R \H \h \V \v \X \Z \z $ $M */
721 { 0, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0 }, /* \D */
722 { 1, 0, 0, 1, 1, 0, 0, 0, 0, 0, 0, 1, 0, 1, 0, 1, 0, 1, 1, 1, 1 }, /* \d */
723 { 0, 0, 0, 1, 0, 0, 0, 0, 0, 0, 0, 1, 0, 1, 0, 1, 0, 1, 1, 1, 1 }, /* \S */
724 { 0, 1, 1, 0, 0, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0 }, /* \s */
725 { 0, 1, 0, 0, 0, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0 }, /* \W */
726 { 0, 0, 0, 1, 1, 0, 0, 0, 0, 0, 0, 1, 0, 1, 0, 1, 0, 1, 1, 1, 1 }, /* \w */
727 { 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0, 0, 0, 0, 0, 1, 0, 0 }, /* . */
728 { 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0 }, /* .+ */
729 { 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0 }, /* \C */
730 { 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0 }, /* \P */
731 { 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0 }, /* \p */
732 { 0, 1, 0, 1, 0, 1, 1, 0, 0, 0, 0, 0, 0, 1, 0, 0, 0, 0, 1, 0, 0 }, /* \R */
733 { 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0, 0, 0, 1, 0, 0 }, /* \H */
734 { 0, 1, 1, 0, 0, 1, 0, 0, 0, 0, 0, 1, 1, 0, 0, 1, 0, 0, 1, 0, 0 }, /* \h */
735 { 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0, 0, 1, 0, 0, 1, 0, 0 }, /* \V */
736 { 0, 1, 1, 0, 0, 1, 0, 0, 0, 0, 0, 0, 0, 1, 1, 0, 0, 0, 1, 0, 0 }, /* \v */
737 { 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0 } /* \X */
738 };
739
740
741 /* This table is used to check whether auto-possessification is possible
742 between adjacent Unicode property opcodes (OP_PROP and OP_NOTPROP). The
743 left-hand (repeated) opcode is used to select the row, and the right-hand
744 opcode is used to select the column. The values are as follows:
745
746 0 Always return FALSE (never auto-possessify)
747 1 Character groups are distinct (possessify if both are OP_PROP)
748 2 Check character categories in the same group (general or particular)
749 3 TRUE if the two opcodes are not the same (PROP vs NOTPROP)
750
751 4 Check left general category vs right particular category
752 5 Check right general category vs left particular category
753
754 6 Left alphanum vs right general category
755 7 Left space vs right general category
756 8 Left word vs right general category
757
758 9 Right alphanum vs left general category
759 10 Right space vs left general category
760 11 Right word vs left general category
761
762 12 Left alphanum vs right particular category
763 13 Left space vs right particular category
764 14 Left word vs right particular category
765
766 15 Right alphanum vs left particular category
767 16 Right space vs left particular category
768 17 Right word vs left particular category
769 */
770
771 static const pcre_uint8 propposstab[PT_TABSIZE][PT_TABSIZE] = {
772 /* ANY LAMP GC PC SC ALNUM SPACE PXSPACE WORD CLIST UCNC */
773 { 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0 }, /* PT_ANY */
774 { 0, 3, 0, 0, 0, 3, 1, 1, 0, 0, 0 }, /* PT_LAMP */
775 { 0, 0, 2, 4, 0, 9, 10, 10, 11, 0, 0 }, /* PT_GC */
776 { 0, 0, 5, 2, 0, 15, 16, 16, 17, 0, 0 }, /* PT_PC */
777 { 0, 0, 0, 0, 2, 0, 0, 0, 0, 0, 0 }, /* PT_SC */
778 { 0, 3, 6, 12, 0, 3, 1, 1, 0, 0, 0 }, /* PT_ALNUM */
779 { 0, 1, 7, 13, 0, 1, 3, 3, 1, 0, 0 }, /* PT_SPACE */
780 { 0, 1, 7, 13, 0, 1, 3, 3, 1, 0, 0 }, /* PT_PXSPACE */
781 { 0, 0, 8, 14, 0, 0, 1, 1, 3, 0, 0 }, /* PT_WORD */
782 { 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0 }, /* PT_CLIST */
783 { 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 3 } /* PT_UCNC */
784 };
785
786 /* This table is used to check whether auto-possessification is possible
787 between adjacent Unicode property opcodes (OP_PROP and OP_NOTPROP) when one
788 specifies a general category and the other specifies a particular category. The
789 row is selected by the general category and the column by the particular
790 category. The value is 1 if the particular category is not part of the general
791 category. */
792
793 static const pcre_uint8 catposstab[7][30] = {
794 /* Cc Cf Cn Co Cs Ll Lm Lo Lt Lu Mc Me Mn Nd Nl No Pc Pd Pe Pf Pi Po Ps Sc Sk Sm So Zl Zp Zs */
795 { 0, 0, 0, 0, 0, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1 }, /* C */
796 { 1, 1, 1, 1, 1, 0, 0, 0, 0, 0, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1 }, /* L */
797 { 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 0, 0, 0, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1 }, /* M */
798 { 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 0, 0, 0, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1 }, /* N */
799 { 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 0, 0, 0, 0, 0, 0, 0, 1, 1, 1, 1, 1, 1, 1 }, /* P */
800 { 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 0, 0, 0, 0, 1, 1, 1 }, /* S */
801 { 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 0, 0, 0 } /* Z */
802 };
803
804 /* This table is used when checking ALNUM, (PX)SPACE, SPACE, and WORD against
805 a general or particular category. The properties in each row are those
806 that apply to the character set in question. Duplication means that a little
807 unnecessary work is done when checking, but this keeps things much simpler
808 because they can all use the same code. For more details see the comment where
809 this table is used.
810
811 Note: SPACE and PXSPACE used to be different because Perl excluded VT from
812 "space", but from Perl 5.18 it's included, so both categories are treated the
813 same here. */
814
815 static const pcre_uint8 posspropstab[3][4] = {
816 { ucp_L, ucp_N, ucp_N, ucp_Nl }, /* ALNUM, 3rd and 4th values redundant */
817 { ucp_Z, ucp_Z, ucp_C, ucp_Cc }, /* SPACE and PXSPACE, 2nd value redundant */
818 { ucp_L, ucp_N, ucp_P, ucp_Po } /* WORD */
819 };
820
821 /* This table is used when converting repeating opcodes into possessified
822 versions as a result of an explicit possessive quantifier such as ++. A zero
823 value means there is no possessified version - in those cases the item in
824 question must be wrapped in ONCE brackets. The table is truncated at OP_CALLOUT
825 because all relevant opcodes are less than that. */
826
827 static const pcre_uint8 opcode_possessify[] = {
828 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, /* 0 - 15 */
829 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, /* 16 - 31 */
830
831 0, /* NOTI */
832 OP_POSSTAR, 0, /* STAR, MINSTAR */
833 OP_POSPLUS, 0, /* PLUS, MINPLUS */
834 OP_POSQUERY, 0, /* QUERY, MINQUERY */
835 OP_POSUPTO, 0, /* UPTO, MINUPTO */
836 0, /* EXACT */
837 0, 0, 0, 0, /* POS{STAR,PLUS,QUERY,UPTO} */
838
839 OP_POSSTARI, 0, /* STARI, MINSTARI */
840 OP_POSPLUSI, 0, /* PLUSI, MINPLUSI */
841 OP_POSQUERYI, 0, /* QUERYI, MINQUERYI */
842 OP_POSUPTOI, 0, /* UPTOI, MINUPTOI */
843 0, /* EXACTI */
844 0, 0, 0, 0, /* POS{STARI,PLUSI,QUERYI,UPTOI} */
845
846 OP_NOTPOSSTAR, 0, /* NOTSTAR, NOTMINSTAR */
847 OP_NOTPOSPLUS, 0, /* NOTPLUS, NOTMINPLUS */
848 OP_NOTPOSQUERY, 0, /* NOTQUERY, NOTMINQUERY */
849 OP_NOTPOSUPTO, 0, /* NOTUPTO, NOTMINUPTO */
850 0, /* NOTEXACT */
851 0, 0, 0, 0, /* NOTPOS{STAR,PLUS,QUERY,UPTO} */
852
853 OP_NOTPOSSTARI, 0, /* NOTSTARI, NOTMINSTARI */
854 OP_NOTPOSPLUSI, 0, /* NOTPLUSI, NOTMINPLUSI */
855 OP_NOTPOSQUERYI, 0, /* NOTQUERYI, NOTMINQUERYI */
856 OP_NOTPOSUPTOI, 0, /* NOTUPTOI, NOTMINUPTOI */
857 0, /* NOTEXACTI */
858 0, 0, 0, 0, /* NOTPOS{STARI,PLUSI,QUERYI,UPTOI} */
859
860 OP_TYPEPOSSTAR, 0, /* TYPESTAR, TYPEMINSTAR */
861 OP_TYPEPOSPLUS, 0, /* TYPEPLUS, TYPEMINPLUS */
862 OP_TYPEPOSQUERY, 0, /* TYPEQUERY, TYPEMINQUERY */
863 OP_TYPEPOSUPTO, 0, /* TYPEUPTO, TYPEMINUPTO */
864 0, /* TYPEEXACT */
865 0, 0, 0, 0, /* TYPEPOS{STAR,PLUS,QUERY,UPTO} */
866
867 OP_CRPOSSTAR, 0, /* CRSTAR, CRMINSTAR */
868 OP_CRPOSPLUS, 0, /* CRPLUS, CRMINPLUS */
869 OP_CRPOSQUERY, 0, /* CRQUERY, CRMINQUERY */
870 OP_CRPOSRANGE, 0, /* CRRANGE, CRMINRANGE */
871 0, 0, 0, 0, /* CRPOS{STAR,PLUS,QUERY,RANGE} */
872
873 0, 0, 0, /* CLASS, NCLASS, XCLASS */
874 0, 0, /* REF, REFI */
875 0, 0, /* DNREF, DNREFI */
876 0, 0 /* RECURSE, CALLOUT */
877 };
878
879
880
881 /*************************************************
882 * Find an error text *
883 *************************************************/
884
885 /* The error texts are now all in one long string, to save on relocations. As
886 some of the text is of unknown length, we can't use a table of offsets.
887 Instead, just count through the strings. This is not a performance issue
888 because it happens only when there has been a compilation error.
889
890 Argument: the error number
891 Returns: pointer to the error string
892 */
893
894 static const char *
find_error_text(int n)895 find_error_text(int n)
896 {
897 const char *s = error_texts;
898 for (; n > 0; n--)
899 {
900 while (*s++ != CHAR_NULL) {};
901 if (*s == CHAR_NULL) return "Error text not found (please report)";
902 }
903 return s;
904 }
905
906
907
908 /*************************************************
909 * Expand the workspace *
910 *************************************************/
911
912 /* This function is called during the second compiling phase, if the number of
913 forward references fills the existing workspace, which is originally a block on
914 the stack. A larger block is obtained from malloc() unless the ultimate limit
915 has been reached or the increase will be rather small.
916
917 Argument: pointer to the compile data block
918 Returns: 0 if all went well, else an error number
919 */
920
921 static int
expand_workspace(compile_data * cd)922 expand_workspace(compile_data *cd)
923 {
924 pcre_uchar *newspace;
925 int newsize = cd->workspace_size * 2;
926
927 if (newsize > COMPILE_WORK_SIZE_MAX) newsize = COMPILE_WORK_SIZE_MAX;
928 if (cd->workspace_size >= COMPILE_WORK_SIZE_MAX ||
929 newsize - cd->workspace_size < WORK_SIZE_SAFETY_MARGIN)
930 return ERR72;
931
932 newspace = (PUBL(malloc))(IN_UCHARS(newsize));
933 if (newspace == NULL) return ERR21;
934 memcpy(newspace, cd->start_workspace, cd->workspace_size * sizeof(pcre_uchar));
935 cd->hwm = (pcre_uchar *)newspace + (cd->hwm - cd->start_workspace);
936 if (cd->workspace_size > COMPILE_WORK_SIZE)
937 (PUBL(free))((void *)cd->start_workspace);
938 cd->start_workspace = newspace;
939 cd->workspace_size = newsize;
940 return 0;
941 }
942
943
944
945 /*************************************************
946 * Check for counted repeat *
947 *************************************************/
948
949 /* This function is called when a '{' is encountered in a place where it might
950 start a quantifier. It looks ahead to see if it really is a quantifier or not.
951 It is only a quantifier if it is one of the forms {ddd} {ddd,} or {ddd,ddd}
952 where the ddds are digits.
953
954 Arguments:
955 p pointer to the first char after '{'
956
957 Returns: TRUE or FALSE
958 */
959
960 static BOOL
is_counted_repeat(const pcre_uchar * p)961 is_counted_repeat(const pcre_uchar *p)
962 {
963 if (!IS_DIGIT(*p)) return FALSE;
964 p++;
965 while (IS_DIGIT(*p)) p++;
966 if (*p == CHAR_RIGHT_CURLY_BRACKET) return TRUE;
967
968 if (*p++ != CHAR_COMMA) return FALSE;
969 if (*p == CHAR_RIGHT_CURLY_BRACKET) return TRUE;
970
971 if (!IS_DIGIT(*p)) return FALSE;
972 p++;
973 while (IS_DIGIT(*p)) p++;
974
975 return (*p == CHAR_RIGHT_CURLY_BRACKET);
976 }
977
978
979
980 /*************************************************
981 * Handle escapes *
982 *************************************************/
983
984 /* This function is called when a \ has been encountered. It either returns a
985 positive value for a simple escape such as \n, or 0 for a data character which
986 will be placed in chptr. A backreference to group n is returned as negative n.
987 When UTF-8 is enabled, a positive value greater than 255 may be returned in
988 chptr. On entry, ptr is pointing at the \. On exit, it is on the final
989 character of the escape sequence.
990
991 Arguments:
992 ptrptr points to the pattern position pointer
993 chptr points to a returned data character
994 errorcodeptr points to the errorcode variable
995 bracount number of previous extracting brackets
996 options the options bits
997 isclass TRUE if inside a character class
998
999 Returns: zero => a data character
1000 positive => a special escape sequence
1001 negative => a back reference
1002 on error, errorcodeptr is set
1003 */
1004
1005 static int
check_escape(const pcre_uchar ** ptrptr,pcre_uint32 * chptr,int * errorcodeptr,int bracount,int options,BOOL isclass)1006 check_escape(const pcre_uchar **ptrptr, pcre_uint32 *chptr, int *errorcodeptr,
1007 int bracount, int options, BOOL isclass)
1008 {
1009 /* PCRE_UTF16 has the same value as PCRE_UTF8. */
1010 BOOL utf = (options & PCRE_UTF8) != 0;
1011 const pcre_uchar *ptr = *ptrptr + 1;
1012 pcre_uint32 c;
1013 int escape = 0;
1014 int i;
1015
1016 GETCHARINCTEST(c, ptr); /* Get character value, increment pointer */
1017 ptr--; /* Set pointer back to the last byte */
1018
1019 /* If backslash is at the end of the pattern, it's an error. */
1020
1021 if (c == CHAR_NULL) *errorcodeptr = ERR1;
1022
1023 /* Non-alphanumerics are literals. For digits or letters, do an initial lookup
1024 in a table. A non-zero result is something that can be returned immediately.
1025 Otherwise further processing may be required. */
1026
1027 #ifndef EBCDIC /* ASCII/UTF-8 coding */
1028 /* Not alphanumeric */
1029 else if (c < CHAR_0 || c > CHAR_z) {}
1030 else if ((i = escapes[c - CHAR_0]) != 0)
1031 { if (i > 0) c = (pcre_uint32)i; else escape = -i; }
1032
1033 #else /* EBCDIC coding */
1034 /* Not alphanumeric */
1035 else if (c < CHAR_a || (!MAX_255(c) || (ebcdic_chartab[c] & 0x0E) == 0)) {}
1036 else if ((i = escapes[c - 0x48]) != 0) { if (i > 0) c = (pcre_uint32)i; else escape = -i; }
1037 #endif
1038
1039 /* Escapes that need further processing, or are illegal. */
1040
1041 else
1042 {
1043 const pcre_uchar *oldptr;
1044 BOOL braced, negated, overflow;
1045 int s;
1046
1047 switch (c)
1048 {
1049 /* A number of Perl escapes are not handled by PCRE. We give an explicit
1050 error. */
1051
1052 case CHAR_l:
1053 case CHAR_L:
1054 *errorcodeptr = ERR37;
1055 break;
1056
1057 case CHAR_u:
1058 if ((options & PCRE_JAVASCRIPT_COMPAT) != 0)
1059 {
1060 /* In JavaScript, \u must be followed by four hexadecimal numbers.
1061 Otherwise it is a lowercase u letter. */
1062 if (MAX_255(ptr[1]) && (digitab[ptr[1]] & ctype_xdigit) != 0
1063 && MAX_255(ptr[2]) && (digitab[ptr[2]] & ctype_xdigit) != 0
1064 && MAX_255(ptr[3]) && (digitab[ptr[3]] & ctype_xdigit) != 0
1065 && MAX_255(ptr[4]) && (digitab[ptr[4]] & ctype_xdigit) != 0)
1066 {
1067 c = 0;
1068 for (i = 0; i < 4; ++i)
1069 {
1070 register pcre_uint32 cc = *(++ptr);
1071 #ifndef EBCDIC /* ASCII/UTF-8 coding */
1072 if (cc >= CHAR_a) cc -= 32; /* Convert to upper case */
1073 c = (c << 4) + cc - ((cc < CHAR_A)? CHAR_0 : (CHAR_A - 10));
1074 #else /* EBCDIC coding */
1075 if (cc >= CHAR_a && cc <= CHAR_z) cc += 64; /* Convert to upper case */
1076 c = (c << 4) + cc - ((cc >= CHAR_0)? CHAR_0 : (CHAR_A - 10));
1077 #endif
1078 }
1079
1080 #if defined COMPILE_PCRE8
1081 if (c > (utf ? 0x10ffffU : 0xffU))
1082 #elif defined COMPILE_PCRE16
1083 if (c > (utf ? 0x10ffffU : 0xffffU))
1084 #elif defined COMPILE_PCRE32
1085 if (utf && c > 0x10ffffU)
1086 #endif
1087 {
1088 *errorcodeptr = ERR76;
1089 }
1090 else if (utf && c >= 0xd800 && c <= 0xdfff) *errorcodeptr = ERR73;
1091 }
1092 }
1093 else
1094 *errorcodeptr = ERR37;
1095 break;
1096
1097 case CHAR_U:
1098 /* In JavaScript, \U is an uppercase U letter. */
1099 if ((options & PCRE_JAVASCRIPT_COMPAT) == 0) *errorcodeptr = ERR37;
1100 break;
1101
1102 /* In a character class, \g is just a literal "g". Outside a character
1103 class, \g must be followed by one of a number of specific things:
1104
1105 (1) A number, either plain or braced. If positive, it is an absolute
1106 backreference. If negative, it is a relative backreference. This is a Perl
1107 5.10 feature.
1108
1109 (2) Perl 5.10 also supports \g{name} as a reference to a named group. This
1110 is part of Perl's movement towards a unified syntax for back references. As
1111 this is synonymous with \k{name}, we fudge it up by pretending it really
1112 was \k.
1113
1114 (3) For Oniguruma compatibility we also support \g followed by a name or a
1115 number either in angle brackets or in single quotes. However, these are
1116 (possibly recursive) subroutine calls, _not_ backreferences. Just return
1117 the ESC_g code (cf \k). */
1118
1119 case CHAR_g:
1120 if (isclass) break;
1121 if (ptr[1] == CHAR_LESS_THAN_SIGN || ptr[1] == CHAR_APOSTROPHE)
1122 {
1123 escape = ESC_g;
1124 break;
1125 }
1126
1127 /* Handle the Perl-compatible cases */
1128
1129 if (ptr[1] == CHAR_LEFT_CURLY_BRACKET)
1130 {
1131 const pcre_uchar *p;
1132 for (p = ptr+2; *p != CHAR_NULL && *p != CHAR_RIGHT_CURLY_BRACKET; p++)
1133 if (*p != CHAR_MINUS && !IS_DIGIT(*p)) break;
1134 if (*p != CHAR_NULL && *p != CHAR_RIGHT_CURLY_BRACKET)
1135 {
1136 escape = ESC_k;
1137 break;
1138 }
1139 braced = TRUE;
1140 ptr++;
1141 }
1142 else braced = FALSE;
1143
1144 if (ptr[1] == CHAR_MINUS)
1145 {
1146 negated = TRUE;
1147 ptr++;
1148 }
1149 else negated = FALSE;
1150
1151 /* The integer range is limited by the machine's int representation. */
1152 s = 0;
1153 overflow = FALSE;
1154 while (IS_DIGIT(ptr[1]))
1155 {
1156 if (s > INT_MAX / 10 - 1) /* Integer overflow */
1157 {
1158 overflow = TRUE;
1159 break;
1160 }
1161 s = s * 10 + (int)(*(++ptr) - CHAR_0);
1162 }
1163 if (overflow) /* Integer overflow */
1164 {
1165 while (IS_DIGIT(ptr[1]))
1166 ptr++;
1167 *errorcodeptr = ERR61;
1168 break;
1169 }
1170
1171 if (braced && *(++ptr) != CHAR_RIGHT_CURLY_BRACKET)
1172 {
1173 *errorcodeptr = ERR57;
1174 break;
1175 }
1176
1177 if (s == 0)
1178 {
1179 *errorcodeptr = ERR58;
1180 break;
1181 }
1182
1183 if (negated)
1184 {
1185 if (s > bracount)
1186 {
1187 *errorcodeptr = ERR15;
1188 break;
1189 }
1190 s = bracount - (s - 1);
1191 }
1192
1193 escape = -s;
1194 break;
1195
1196 /* The handling of escape sequences consisting of a string of digits
1197 starting with one that is not zero is not straightforward. Perl has changed
1198 over the years. Nowadays \g{} for backreferences and \o{} for octal are
1199 recommended to avoid the ambiguities in the old syntax.
1200
1201 Outside a character class, the digits are read as a decimal number. If the
1202 number is less than 8 (used to be 10), or if there are that many previous
1203 extracting left brackets, then it is a back reference. Otherwise, up to
1204 three octal digits are read to form an escaped byte. Thus \123 is likely to
1205 be octal 123 (cf \0123, which is octal 012 followed by the literal 3). If
1206 the octal value is greater than 377, the least significant 8 bits are
1207 taken. \8 and \9 are treated as the literal characters 8 and 9.
1208
1209 Inside a character class, \ followed by a digit is always either a literal
1210 8 or 9 or an octal number. */
1211
1212 case CHAR_1: case CHAR_2: case CHAR_3: case CHAR_4: case CHAR_5:
1213 case CHAR_6: case CHAR_7: case CHAR_8: case CHAR_9:
1214
1215 if (!isclass)
1216 {
1217 oldptr = ptr;
1218 /* The integer range is limited by the machine's int representation. */
1219 s = (int)(c -CHAR_0);
1220 overflow = FALSE;
1221 while (IS_DIGIT(ptr[1]))
1222 {
1223 if (s > INT_MAX / 10 - 1) /* Integer overflow */
1224 {
1225 overflow = TRUE;
1226 break;
1227 }
1228 s = s * 10 + (int)(*(++ptr) - CHAR_0);
1229 }
1230 if (overflow) /* Integer overflow */
1231 {
1232 while (IS_DIGIT(ptr[1]))
1233 ptr++;
1234 *errorcodeptr = ERR61;
1235 break;
1236 }
1237 if (s < 8 || s <= bracount) /* Check for back reference */
1238 {
1239 escape = -s;
1240 break;
1241 }
1242 ptr = oldptr; /* Put the pointer back and fall through */
1243 }
1244
1245 /* Handle a digit following \ when the number is not a back reference. If
1246 the first digit is 8 or 9, Perl used to generate a binary zero byte and
1247 then treat the digit as a following literal. At least by Perl 5.18 this
1248 changed so as not to insert the binary zero. */
1249
1250 if ((c = *ptr) >= CHAR_8) break;
1251
1252 /* fall through */
1253 /* Fall through with a digit less than 8 */
1254
1255 /* \0 always starts an octal number, but we may drop through to here with a
1256 larger first octal digit. The original code used just to take the least
1257 significant 8 bits of octal numbers (I think this is what early Perls used
1258 to do). Nowadays we allow for larger numbers in UTF-8 mode and 16-bit mode,
1259 but no more than 3 octal digits. */
1260
1261 case CHAR_0:
1262 c -= CHAR_0;
1263 while(i++ < 2 && ptr[1] >= CHAR_0 && ptr[1] <= CHAR_7)
1264 c = c * 8 + *(++ptr) - CHAR_0;
1265 #ifdef COMPILE_PCRE8
1266 if (!utf && c > 0xff) *errorcodeptr = ERR51;
1267 #endif
1268 break;
1269
1270 /* \o is a relatively new Perl feature, supporting a more general way of
1271 specifying character codes in octal. The only supported form is \o{ddd}. */
1272
1273 case CHAR_o:
1274 if (ptr[1] != CHAR_LEFT_CURLY_BRACKET) *errorcodeptr = ERR81; else
1275 if (ptr[2] == CHAR_RIGHT_CURLY_BRACKET) *errorcodeptr = ERR86; else
1276 {
1277 ptr += 2;
1278 c = 0;
1279 overflow = FALSE;
1280 while (*ptr >= CHAR_0 && *ptr <= CHAR_7)
1281 {
1282 register pcre_uint32 cc = *ptr++;
1283 if (c == 0 && cc == CHAR_0) continue; /* Leading zeroes */
1284 #ifdef COMPILE_PCRE32
1285 if (c >= 0x20000000l) { overflow = TRUE; break; }
1286 #endif
1287 c = (c << 3) + cc - CHAR_0 ;
1288 #if defined COMPILE_PCRE8
1289 if (c > (utf ? 0x10ffffU : 0xffU)) { overflow = TRUE; break; }
1290 #elif defined COMPILE_PCRE16
1291 if (c > (utf ? 0x10ffffU : 0xffffU)) { overflow = TRUE; break; }
1292 #elif defined COMPILE_PCRE32
1293 if (utf && c > 0x10ffffU) { overflow = TRUE; break; }
1294 #endif
1295 }
1296 if (overflow)
1297 {
1298 while (*ptr >= CHAR_0 && *ptr <= CHAR_7) ptr++;
1299 *errorcodeptr = ERR34;
1300 }
1301 else if (*ptr == CHAR_RIGHT_CURLY_BRACKET)
1302 {
1303 if (utf && c >= 0xd800 && c <= 0xdfff) *errorcodeptr = ERR73;
1304 }
1305 else *errorcodeptr = ERR80;
1306 }
1307 break;
1308
1309 /* \x is complicated. In JavaScript, \x must be followed by two hexadecimal
1310 numbers. Otherwise it is a lowercase x letter. */
1311
1312 case CHAR_x:
1313 if ((options & PCRE_JAVASCRIPT_COMPAT) != 0)
1314 {
1315 if (MAX_255(ptr[1]) && (digitab[ptr[1]] & ctype_xdigit) != 0
1316 && MAX_255(ptr[2]) && (digitab[ptr[2]] & ctype_xdigit) != 0)
1317 {
1318 c = 0;
1319 for (i = 0; i < 2; ++i)
1320 {
1321 register pcre_uint32 cc = *(++ptr);
1322 #ifndef EBCDIC /* ASCII/UTF-8 coding */
1323 if (cc >= CHAR_a) cc -= 32; /* Convert to upper case */
1324 c = (c << 4) + cc - ((cc < CHAR_A)? CHAR_0 : (CHAR_A - 10));
1325 #else /* EBCDIC coding */
1326 if (cc >= CHAR_a && cc <= CHAR_z) cc += 64; /* Convert to upper case */
1327 c = (c << 4) + cc - ((cc >= CHAR_0)? CHAR_0 : (CHAR_A - 10));
1328 #endif
1329 }
1330 }
1331 } /* End JavaScript handling */
1332
1333 /* Handle \x in Perl's style. \x{ddd} is a character number which can be
1334 greater than 0xff in utf or non-8bit mode, but only if the ddd are hex
1335 digits. If not, { used to be treated as a data character. However, Perl
1336 seems to read hex digits up to the first non-such, and ignore the rest, so
1337 that, for example \x{zz} matches a binary zero. This seems crazy, so PCRE
1338 now gives an error. */
1339
1340 else
1341 {
1342 if (ptr[1] == CHAR_LEFT_CURLY_BRACKET)
1343 {
1344 ptr += 2;
1345 if (*ptr == CHAR_RIGHT_CURLY_BRACKET)
1346 {
1347 *errorcodeptr = ERR86;
1348 break;
1349 }
1350 c = 0;
1351 overflow = FALSE;
1352 while (MAX_255(*ptr) && (digitab[*ptr] & ctype_xdigit) != 0)
1353 {
1354 register pcre_uint32 cc = *ptr++;
1355 if (c == 0 && cc == CHAR_0) continue; /* Leading zeroes */
1356
1357 #ifdef COMPILE_PCRE32
1358 if (c >= 0x10000000l) { overflow = TRUE; break; }
1359 #endif
1360
1361 #ifndef EBCDIC /* ASCII/UTF-8 coding */
1362 if (cc >= CHAR_a) cc -= 32; /* Convert to upper case */
1363 c = (c << 4) + cc - ((cc < CHAR_A)? CHAR_0 : (CHAR_A - 10));
1364 #else /* EBCDIC coding */
1365 if (cc >= CHAR_a && cc <= CHAR_z) cc += 64; /* Convert to upper case */
1366 c = (c << 4) + cc - ((cc >= CHAR_0)? CHAR_0 : (CHAR_A - 10));
1367 #endif
1368
1369 #if defined COMPILE_PCRE8
1370 if (c > (utf ? 0x10ffffU : 0xffU)) { overflow = TRUE; break; }
1371 #elif defined COMPILE_PCRE16
1372 if (c > (utf ? 0x10ffffU : 0xffffU)) { overflow = TRUE; break; }
1373 #elif defined COMPILE_PCRE32
1374 if (utf && c > 0x10ffffU) { overflow = TRUE; break; }
1375 #endif
1376 }
1377
1378 if (overflow)
1379 {
1380 while (MAX_255(*ptr) && (digitab[*ptr] & ctype_xdigit) != 0) ptr++;
1381 *errorcodeptr = ERR34;
1382 }
1383
1384 else if (*ptr == CHAR_RIGHT_CURLY_BRACKET)
1385 {
1386 if (utf && c >= 0xd800 && c <= 0xdfff) *errorcodeptr = ERR73;
1387 }
1388
1389 /* If the sequence of hex digits does not end with '}', give an error.
1390 We used just to recognize this construct and fall through to the normal
1391 \x handling, but nowadays Perl gives an error, which seems much more
1392 sensible, so we do too. */
1393
1394 else *errorcodeptr = ERR79;
1395 } /* End of \x{} processing */
1396
1397 /* Read a single-byte hex-defined char (up to two hex digits after \x) */
1398
1399 else
1400 {
1401 c = 0;
1402 while (i++ < 2 && MAX_255(ptr[1]) && (digitab[ptr[1]] & ctype_xdigit) != 0)
1403 {
1404 pcre_uint32 cc; /* Some compilers don't like */
1405 cc = *(++ptr); /* ++ in initializers */
1406 #ifndef EBCDIC /* ASCII/UTF-8 coding */
1407 if (cc >= CHAR_a) cc -= 32; /* Convert to upper case */
1408 c = c * 16 + cc - ((cc < CHAR_A)? CHAR_0 : (CHAR_A - 10));
1409 #else /* EBCDIC coding */
1410 if (cc <= CHAR_z) cc += 64; /* Convert to upper case */
1411 c = c * 16 + cc - ((cc >= CHAR_0)? CHAR_0 : (CHAR_A - 10));
1412 #endif
1413 }
1414 } /* End of \xdd handling */
1415 } /* End of Perl-style \x handling */
1416 break;
1417
1418 /* For \c, a following letter is upper-cased; then the 0x40 bit is flipped.
1419 An error is given if the byte following \c is not an ASCII character. This
1420 coding is ASCII-specific, but then the whole concept of \cx is
1421 ASCII-specific. (However, an EBCDIC equivalent has now been added.) */
1422
1423 case CHAR_c:
1424 c = *(++ptr);
1425 if (c == CHAR_NULL)
1426 {
1427 *errorcodeptr = ERR2;
1428 break;
1429 }
1430 #ifndef EBCDIC /* ASCII/UTF-8 coding */
1431 if (c > 127) /* Excludes all non-ASCII in either mode */
1432 {
1433 *errorcodeptr = ERR68;
1434 break;
1435 }
1436 if (c >= CHAR_a && c <= CHAR_z) c -= 32;
1437 c ^= 0x40;
1438 #else /* EBCDIC coding */
1439 if (c >= CHAR_a && c <= CHAR_z) c += 64;
1440 if (c == CHAR_QUESTION_MARK)
1441 c = ('\\' == 188 && '`' == 74)? 0x5f : 0xff;
1442 else
1443 {
1444 for (i = 0; i < 32; i++)
1445 {
1446 if (c == ebcdic_escape_c[i]) break;
1447 }
1448 if (i < 32) c = i; else *errorcodeptr = ERR68;
1449 }
1450 #endif
1451 break;
1452
1453 /* PCRE_EXTRA enables extensions to Perl in the matter of escapes. Any
1454 other alphanumeric following \ is an error if PCRE_EXTRA was set;
1455 otherwise, for Perl compatibility, it is a literal. This code looks a bit
1456 odd, but there used to be some cases other than the default, and there may
1457 be again in future, so I haven't "optimized" it. */
1458
1459 default:
1460 if ((options & PCRE_EXTRA) != 0) switch(c)
1461 {
1462 default:
1463 *errorcodeptr = ERR3;
1464 break;
1465 }
1466 break;
1467 }
1468 }
1469
1470 /* Perl supports \N{name} for character names, as well as plain \N for "not
1471 newline". PCRE does not support \N{name}. However, it does support
1472 quantification such as \N{2,3}. */
1473
1474 if (escape == ESC_N && ptr[1] == CHAR_LEFT_CURLY_BRACKET &&
1475 !is_counted_repeat(ptr+2))
1476 *errorcodeptr = ERR37;
1477
1478 /* If PCRE_UCP is set, we change the values for \d etc. */
1479
1480 if ((options & PCRE_UCP) != 0 && escape >= ESC_D && escape <= ESC_w)
1481 escape += (ESC_DU - ESC_D);
1482
1483 /* Set the pointer to the final character before returning. */
1484
1485 *ptrptr = ptr;
1486 *chptr = c;
1487 return escape;
1488 }
1489
1490
1491
1492 #ifdef SUPPORT_UCP
1493 /*************************************************
1494 * Handle \P and \p *
1495 *************************************************/
1496
1497 /* This function is called after \P or \p has been encountered, provided that
1498 PCRE is compiled with support for Unicode properties. On entry, ptrptr is
1499 pointing at the P or p. On exit, it is pointing at the final character of the
1500 escape sequence.
1501
1502 Argument:
1503 ptrptr points to the pattern position pointer
1504 negptr points to a boolean that is set TRUE for negation else FALSE
1505 ptypeptr points to an unsigned int that is set to the type value
1506 pdataptr points to an unsigned int that is set to the detailed property value
1507 errorcodeptr points to the error code variable
1508
1509 Returns: TRUE if the type value was found, or FALSE for an invalid type
1510 */
1511
1512 static BOOL
get_ucp(const pcre_uchar ** ptrptr,BOOL * negptr,unsigned int * ptypeptr,unsigned int * pdataptr,int * errorcodeptr)1513 get_ucp(const pcre_uchar **ptrptr, BOOL *negptr, unsigned int *ptypeptr,
1514 unsigned int *pdataptr, int *errorcodeptr)
1515 {
1516 pcre_uchar c;
1517 int i, bot, top;
1518 const pcre_uchar *ptr = *ptrptr;
1519 pcre_uchar name[32];
1520
1521 c = *(++ptr);
1522 if (c == CHAR_NULL) goto ERROR_RETURN;
1523
1524 *negptr = FALSE;
1525
1526 /* \P or \p can be followed by a name in {}, optionally preceded by ^ for
1527 negation. */
1528
1529 if (c == CHAR_LEFT_CURLY_BRACKET)
1530 {
1531 if (ptr[1] == CHAR_CIRCUMFLEX_ACCENT)
1532 {
1533 *negptr = TRUE;
1534 ptr++;
1535 }
1536 for (i = 0; i < (int)(sizeof(name) / sizeof(pcre_uchar)) - 1; i++)
1537 {
1538 c = *(++ptr);
1539 if (c == CHAR_NULL) goto ERROR_RETURN;
1540 if (c == CHAR_RIGHT_CURLY_BRACKET) break;
1541 name[i] = c;
1542 }
1543 if (c != CHAR_RIGHT_CURLY_BRACKET) goto ERROR_RETURN;
1544 name[i] = 0;
1545 }
1546
1547 /* Otherwise there is just one following character */
1548
1549 else
1550 {
1551 name[0] = c;
1552 name[1] = 0;
1553 }
1554
1555 *ptrptr = ptr;
1556
1557 /* Search for a recognized property name using binary chop */
1558
1559 bot = 0;
1560 top = PRIV(utt_size);
1561
1562 while (bot < top)
1563 {
1564 int r;
1565 i = (bot + top) >> 1;
1566 r = STRCMP_UC_C8(name, PRIV(utt_names) + PRIV(utt)[i].name_offset);
1567 if (r == 0)
1568 {
1569 *ptypeptr = PRIV(utt)[i].type;
1570 *pdataptr = PRIV(utt)[i].value;
1571 return TRUE;
1572 }
1573 if (r > 0) bot = i + 1; else top = i;
1574 }
1575
1576 *errorcodeptr = ERR47;
1577 *ptrptr = ptr;
1578 return FALSE;
1579
1580 ERROR_RETURN:
1581 *errorcodeptr = ERR46;
1582 *ptrptr = ptr;
1583 return FALSE;
1584 }
1585 #endif
1586
1587
1588
1589 /*************************************************
1590 * Read repeat counts *
1591 *************************************************/
1592
1593 /* Read an item of the form {n,m} and return the values. This is called only
1594 after is_counted_repeat() has confirmed that a repeat-count quantifier exists,
1595 so the syntax is guaranteed to be correct, but we need to check the values.
1596
1597 Arguments:
1598 p pointer to first char after '{'
1599 minp pointer to int for min
1600 maxp pointer to int for max
1601 returned as -1 if no max
1602 errorcodeptr points to error code variable
1603
1604 Returns: pointer to '}' on success;
1605 current ptr on error, with errorcodeptr set non-zero
1606 */
1607
1608 static const pcre_uchar *
read_repeat_counts(const pcre_uchar * p,int * minp,int * maxp,int * errorcodeptr)1609 read_repeat_counts(const pcre_uchar *p, int *minp, int *maxp, int *errorcodeptr)
1610 {
1611 int min = 0;
1612 int max = -1;
1613
1614 while (IS_DIGIT(*p))
1615 {
1616 min = min * 10 + (int)(*p++ - CHAR_0);
1617 if (min > 65535)
1618 {
1619 *errorcodeptr = ERR5;
1620 return p;
1621 }
1622 }
1623
1624 if (*p == CHAR_RIGHT_CURLY_BRACKET) max = min; else
1625 {
1626 if (*(++p) != CHAR_RIGHT_CURLY_BRACKET)
1627 {
1628 max = 0;
1629 while(IS_DIGIT(*p))
1630 {
1631 max = max * 10 + (int)(*p++ - CHAR_0);
1632 if (max > 65535)
1633 {
1634 *errorcodeptr = ERR5;
1635 return p;
1636 }
1637 }
1638 if (max < min)
1639 {
1640 *errorcodeptr = ERR4;
1641 return p;
1642 }
1643 }
1644 }
1645
1646 *minp = min;
1647 *maxp = max;
1648 return p;
1649 }
1650
1651
1652
1653 /*************************************************
1654 * Find first significant op code *
1655 *************************************************/
1656
1657 /* This is called by several functions that scan a compiled expression looking
1658 for a fixed first character, or an anchoring op code etc. It skips over things
1659 that do not influence this. For some calls, it makes sense to skip negative
1660 forward and all backward assertions, and also the \b assertion; for others it
1661 does not.
1662
1663 Arguments:
1664 code pointer to the start of the group
1665 skipassert TRUE if certain assertions are to be skipped
1666
1667 Returns: pointer to the first significant opcode
1668 */
1669
1670 static const pcre_uchar*
first_significant_code(const pcre_uchar * code,BOOL skipassert)1671 first_significant_code(const pcre_uchar *code, BOOL skipassert)
1672 {
1673 for (;;)
1674 {
1675 switch ((int)*code)
1676 {
1677 case OP_ASSERT_NOT:
1678 case OP_ASSERTBACK:
1679 case OP_ASSERTBACK_NOT:
1680 if (!skipassert) return code;
1681 do code += GET(code, 1); while (*code == OP_ALT);
1682 code += PRIV(OP_lengths)[*code];
1683 break;
1684
1685 case OP_WORD_BOUNDARY:
1686 case OP_NOT_WORD_BOUNDARY:
1687 if (!skipassert) return code;
1688 /* Fall through */
1689
1690 case OP_CALLOUT:
1691 case OP_CREF:
1692 case OP_DNCREF:
1693 case OP_RREF:
1694 case OP_DNRREF:
1695 case OP_DEF:
1696 code += PRIV(OP_lengths)[*code];
1697 break;
1698
1699 default:
1700 return code;
1701 }
1702 }
1703 /* Control never reaches here */
1704 }
1705
1706
1707
1708 /*************************************************
1709 * Find the fixed length of a branch *
1710 *************************************************/
1711
1712 /* Scan a branch and compute the fixed length of subject that will match it,
1713 if the length is fixed. This is needed for dealing with backward assertions.
1714 In UTF8 mode, the result is in characters rather than bytes. The branch is
1715 temporarily terminated with OP_END when this function is called.
1716
1717 This function is called when a backward assertion is encountered, so that if it
1718 fails, the error message can point to the correct place in the pattern.
1719 However, we cannot do this when the assertion contains subroutine calls,
1720 because they can be forward references. We solve this by remembering this case
1721 and doing the check at the end; a flag specifies which mode we are running in.
1722
1723 Arguments:
1724 code points to the start of the pattern (the bracket)
1725 utf TRUE in UTF-8 / UTF-16 / UTF-32 mode
1726 atend TRUE if called when the pattern is complete
1727 cd the "compile data" structure
1728 recurses chain of recurse_check to catch mutual recursion
1729
1730 Returns: the fixed length,
1731 or -1 if there is no fixed length,
1732 or -2 if \C was encountered (in UTF-8 mode only)
1733 or -3 if an OP_RECURSE item was encountered and atend is FALSE
1734 or -4 if an unknown opcode was encountered (internal error)
1735 */
1736
1737 static int
find_fixedlength(pcre_uchar * code,BOOL utf,BOOL atend,compile_data * cd,recurse_check * recurses)1738 find_fixedlength(pcre_uchar *code, BOOL utf, BOOL atend, compile_data *cd,
1739 recurse_check *recurses)
1740 {
1741 int length = -1;
1742 recurse_check this_recurse;
1743 register int branchlength = 0;
1744 register pcre_uchar *cc = code + 1 + LINK_SIZE;
1745
1746 /* Scan along the opcodes for this branch. If we get to the end of the
1747 branch, check the length against that of the other branches. */
1748
1749 for (;;)
1750 {
1751 int d;
1752 pcre_uchar *ce, *cs;
1753 register pcre_uchar op = *cc;
1754
1755 switch (op)
1756 {
1757 /* We only need to continue for OP_CBRA (normal capturing bracket) and
1758 OP_BRA (normal non-capturing bracket) because the other variants of these
1759 opcodes are all concerned with unlimited repeated groups, which of course
1760 are not of fixed length. */
1761
1762 case OP_CBRA:
1763 case OP_BRA:
1764 case OP_ONCE:
1765 case OP_ONCE_NC:
1766 case OP_COND:
1767 d = find_fixedlength(cc + ((op == OP_CBRA)? IMM2_SIZE : 0), utf, atend, cd,
1768 recurses);
1769 if (d < 0) return d;
1770 branchlength += d;
1771 do cc += GET(cc, 1); while (*cc == OP_ALT);
1772 cc += 1 + LINK_SIZE;
1773 break;
1774
1775 /* Reached end of a branch; if it's a ket it is the end of a nested call.
1776 If it's ALT it is an alternation in a nested call. An ACCEPT is effectively
1777 an ALT. If it is END it's the end of the outer call. All can be handled by
1778 the same code. Note that we must not include the OP_KETRxxx opcodes here,
1779 because they all imply an unlimited repeat. */
1780
1781 case OP_ALT:
1782 case OP_KET:
1783 case OP_END:
1784 case OP_ACCEPT:
1785 case OP_ASSERT_ACCEPT:
1786 if (length < 0) length = branchlength;
1787 else if (length != branchlength) return -1;
1788 if (*cc != OP_ALT) return length;
1789 cc += 1 + LINK_SIZE;
1790 branchlength = 0;
1791 break;
1792
1793 /* A true recursion implies not fixed length, but a subroutine call may
1794 be OK. If the subroutine is a forward reference, we can't deal with
1795 it until the end of the pattern, so return -3. */
1796
1797 case OP_RECURSE:
1798 if (!atend) return -3;
1799 cs = ce = (pcre_uchar *)cd->start_code + GET(cc, 1); /* Start subpattern */
1800 do ce += GET(ce, 1); while (*ce == OP_ALT); /* End subpattern */
1801 if (cc > cs && cc < ce) return -1; /* Recursion */
1802 else /* Check for mutual recursion */
1803 {
1804 recurse_check *r = recurses;
1805 for (r = recurses; r != NULL; r = r->prev) if (r->group == cs) break;
1806 if (r != NULL) return -1; /* Mutual recursion */
1807 }
1808 this_recurse.prev = recurses;
1809 this_recurse.group = cs;
1810 d = find_fixedlength(cs + IMM2_SIZE, utf, atend, cd, &this_recurse);
1811 if (d < 0) return d;
1812 branchlength += d;
1813 cc += 1 + LINK_SIZE;
1814 break;
1815
1816 /* Skip over assertive subpatterns */
1817
1818 case OP_ASSERT:
1819 case OP_ASSERT_NOT:
1820 case OP_ASSERTBACK:
1821 case OP_ASSERTBACK_NOT:
1822 do cc += GET(cc, 1); while (*cc == OP_ALT);
1823 cc += 1 + LINK_SIZE;
1824 break;
1825
1826 /* Skip over things that don't match chars */
1827
1828 case OP_MARK:
1829 case OP_PRUNE_ARG:
1830 case OP_SKIP_ARG:
1831 case OP_THEN_ARG:
1832 cc += cc[1] + PRIV(OP_lengths)[*cc];
1833 break;
1834
1835 case OP_CALLOUT:
1836 case OP_CIRC:
1837 case OP_CIRCM:
1838 case OP_CLOSE:
1839 case OP_COMMIT:
1840 case OP_CREF:
1841 case OP_DEF:
1842 case OP_DNCREF:
1843 case OP_DNRREF:
1844 case OP_DOLL:
1845 case OP_DOLLM:
1846 case OP_EOD:
1847 case OP_EODN:
1848 case OP_FAIL:
1849 case OP_NOT_WORD_BOUNDARY:
1850 case OP_PRUNE:
1851 case OP_REVERSE:
1852 case OP_RREF:
1853 case OP_SET_SOM:
1854 case OP_SKIP:
1855 case OP_SOD:
1856 case OP_SOM:
1857 case OP_THEN:
1858 case OP_WORD_BOUNDARY:
1859 cc += PRIV(OP_lengths)[*cc];
1860 break;
1861
1862 /* Handle literal characters */
1863
1864 case OP_CHAR:
1865 case OP_CHARI:
1866 case OP_NOT:
1867 case OP_NOTI:
1868 branchlength++;
1869 cc += 2;
1870 #ifdef SUPPORT_UTF
1871 if (utf && HAS_EXTRALEN(cc[-1])) cc += GET_EXTRALEN(cc[-1]);
1872 #endif
1873 break;
1874
1875 /* Handle exact repetitions. The count is already in characters, but we
1876 need to skip over a multibyte character in UTF8 mode. */
1877
1878 case OP_EXACT:
1879 case OP_EXACTI:
1880 case OP_NOTEXACT:
1881 case OP_NOTEXACTI:
1882 branchlength += (int)GET2(cc,1);
1883 cc += 2 + IMM2_SIZE;
1884 #ifdef SUPPORT_UTF
1885 if (utf && HAS_EXTRALEN(cc[-1])) cc += GET_EXTRALEN(cc[-1]);
1886 #endif
1887 break;
1888
1889 case OP_TYPEEXACT:
1890 branchlength += GET2(cc,1);
1891 if (cc[1 + IMM2_SIZE] == OP_PROP || cc[1 + IMM2_SIZE] == OP_NOTPROP)
1892 cc += 2;
1893 cc += 1 + IMM2_SIZE + 1;
1894 break;
1895
1896 /* Handle single-char matchers */
1897
1898 case OP_PROP:
1899 case OP_NOTPROP:
1900 cc += 2;
1901 /* Fall through */
1902
1903 case OP_HSPACE:
1904 case OP_VSPACE:
1905 case OP_NOT_HSPACE:
1906 case OP_NOT_VSPACE:
1907 case OP_NOT_DIGIT:
1908 case OP_DIGIT:
1909 case OP_NOT_WHITESPACE:
1910 case OP_WHITESPACE:
1911 case OP_NOT_WORDCHAR:
1912 case OP_WORDCHAR:
1913 case OP_ANY:
1914 case OP_ALLANY:
1915 branchlength++;
1916 cc++;
1917 break;
1918
1919 /* The single-byte matcher isn't allowed. This only happens in UTF-8 mode;
1920 otherwise \C is coded as OP_ALLANY. */
1921
1922 case OP_ANYBYTE:
1923 return -2;
1924
1925 /* Check a class for variable quantification */
1926
1927 case OP_CLASS:
1928 case OP_NCLASS:
1929 #if defined SUPPORT_UTF || defined COMPILE_PCRE16 || defined COMPILE_PCRE32
1930 case OP_XCLASS:
1931 /* The original code caused an unsigned overflow in 64 bit systems,
1932 so now we use a conditional statement. */
1933 if (op == OP_XCLASS)
1934 cc += GET(cc, 1);
1935 else
1936 cc += PRIV(OP_lengths)[OP_CLASS];
1937 #else
1938 cc += PRIV(OP_lengths)[OP_CLASS];
1939 #endif
1940
1941 switch (*cc)
1942 {
1943 case OP_CRSTAR:
1944 case OP_CRMINSTAR:
1945 case OP_CRPLUS:
1946 case OP_CRMINPLUS:
1947 case OP_CRQUERY:
1948 case OP_CRMINQUERY:
1949 case OP_CRPOSSTAR:
1950 case OP_CRPOSPLUS:
1951 case OP_CRPOSQUERY:
1952 return -1;
1953
1954 case OP_CRRANGE:
1955 case OP_CRMINRANGE:
1956 case OP_CRPOSRANGE:
1957 if (GET2(cc,1) != GET2(cc,1+IMM2_SIZE)) return -1;
1958 branchlength += (int)GET2(cc,1);
1959 cc += 1 + 2 * IMM2_SIZE;
1960 break;
1961
1962 default:
1963 branchlength++;
1964 }
1965 break;
1966
1967 /* Anything else is variable length */
1968
1969 case OP_ANYNL:
1970 case OP_BRAMINZERO:
1971 case OP_BRAPOS:
1972 case OP_BRAPOSZERO:
1973 case OP_BRAZERO:
1974 case OP_CBRAPOS:
1975 case OP_EXTUNI:
1976 case OP_KETRMAX:
1977 case OP_KETRMIN:
1978 case OP_KETRPOS:
1979 case OP_MINPLUS:
1980 case OP_MINPLUSI:
1981 case OP_MINQUERY:
1982 case OP_MINQUERYI:
1983 case OP_MINSTAR:
1984 case OP_MINSTARI:
1985 case OP_MINUPTO:
1986 case OP_MINUPTOI:
1987 case OP_NOTMINPLUS:
1988 case OP_NOTMINPLUSI:
1989 case OP_NOTMINQUERY:
1990 case OP_NOTMINQUERYI:
1991 case OP_NOTMINSTAR:
1992 case OP_NOTMINSTARI:
1993 case OP_NOTMINUPTO:
1994 case OP_NOTMINUPTOI:
1995 case OP_NOTPLUS:
1996 case OP_NOTPLUSI:
1997 case OP_NOTPOSPLUS:
1998 case OP_NOTPOSPLUSI:
1999 case OP_NOTPOSQUERY:
2000 case OP_NOTPOSQUERYI:
2001 case OP_NOTPOSSTAR:
2002 case OP_NOTPOSSTARI:
2003 case OP_NOTPOSUPTO:
2004 case OP_NOTPOSUPTOI:
2005 case OP_NOTQUERY:
2006 case OP_NOTQUERYI:
2007 case OP_NOTSTAR:
2008 case OP_NOTSTARI:
2009 case OP_NOTUPTO:
2010 case OP_NOTUPTOI:
2011 case OP_PLUS:
2012 case OP_PLUSI:
2013 case OP_POSPLUS:
2014 case OP_POSPLUSI:
2015 case OP_POSQUERY:
2016 case OP_POSQUERYI:
2017 case OP_POSSTAR:
2018 case OP_POSSTARI:
2019 case OP_POSUPTO:
2020 case OP_POSUPTOI:
2021 case OP_QUERY:
2022 case OP_QUERYI:
2023 case OP_REF:
2024 case OP_REFI:
2025 case OP_DNREF:
2026 case OP_DNREFI:
2027 case OP_SBRA:
2028 case OP_SBRAPOS:
2029 case OP_SCBRA:
2030 case OP_SCBRAPOS:
2031 case OP_SCOND:
2032 case OP_SKIPZERO:
2033 case OP_STAR:
2034 case OP_STARI:
2035 case OP_TYPEMINPLUS:
2036 case OP_TYPEMINQUERY:
2037 case OP_TYPEMINSTAR:
2038 case OP_TYPEMINUPTO:
2039 case OP_TYPEPLUS:
2040 case OP_TYPEPOSPLUS:
2041 case OP_TYPEPOSQUERY:
2042 case OP_TYPEPOSSTAR:
2043 case OP_TYPEPOSUPTO:
2044 case OP_TYPEQUERY:
2045 case OP_TYPESTAR:
2046 case OP_TYPEUPTO:
2047 case OP_UPTO:
2048 case OP_UPTOI:
2049 return -1;
2050
2051 /* Catch unrecognized opcodes so that when new ones are added they
2052 are not forgotten, as has happened in the past. */
2053
2054 default:
2055 return -4;
2056 }
2057 }
2058 /* Control never gets here */
2059 }
2060
2061
2062
2063 /*************************************************
2064 * Scan compiled regex for specific bracket *
2065 *************************************************/
2066
2067 /* This little function scans through a compiled pattern until it finds a
2068 capturing bracket with the given number, or, if the number is negative, an
2069 instance of OP_REVERSE for a lookbehind. The function is global in the C sense
2070 so that it can be called from pcre_study() when finding the minimum matching
2071 length.
2072
2073 Arguments:
2074 code points to start of expression
2075 utf TRUE in UTF-8 / UTF-16 / UTF-32 mode
2076 number the required bracket number or negative to find a lookbehind
2077
2078 Returns: pointer to the opcode for the bracket, or NULL if not found
2079 */
2080
2081 const pcre_uchar *
PRIV(find_bracket)2082 PRIV(find_bracket)(const pcre_uchar *code, BOOL utf, int number)
2083 {
2084 for (;;)
2085 {
2086 register pcre_uchar c = *code;
2087
2088 if (c == OP_END) return NULL;
2089
2090 /* XCLASS is used for classes that cannot be represented just by a bit
2091 map. This includes negated single high-valued characters. The length in
2092 the table is zero; the actual length is stored in the compiled code. */
2093
2094 if (c == OP_XCLASS) code += GET(code, 1);
2095
2096 /* Handle recursion */
2097
2098 else if (c == OP_REVERSE)
2099 {
2100 if (number < 0) return (pcre_uchar *)code;
2101 code += PRIV(OP_lengths)[c];
2102 }
2103
2104 /* Handle capturing bracket */
2105
2106 else if (c == OP_CBRA || c == OP_SCBRA ||
2107 c == OP_CBRAPOS || c == OP_SCBRAPOS)
2108 {
2109 int n = (int)GET2(code, 1+LINK_SIZE);
2110 if (n == number) return (pcre_uchar *)code;
2111 code += PRIV(OP_lengths)[c];
2112 }
2113
2114 /* Otherwise, we can get the item's length from the table, except that for
2115 repeated character types, we have to test for \p and \P, which have an extra
2116 two bytes of parameters, and for MARK/PRUNE/SKIP/THEN with an argument, we
2117 must add in its length. */
2118
2119 else
2120 {
2121 switch(c)
2122 {
2123 case OP_TYPESTAR:
2124 case OP_TYPEMINSTAR:
2125 case OP_TYPEPLUS:
2126 case OP_TYPEMINPLUS:
2127 case OP_TYPEQUERY:
2128 case OP_TYPEMINQUERY:
2129 case OP_TYPEPOSSTAR:
2130 case OP_TYPEPOSPLUS:
2131 case OP_TYPEPOSQUERY:
2132 if (code[1] == OP_PROP || code[1] == OP_NOTPROP) code += 2;
2133 break;
2134
2135 case OP_TYPEUPTO:
2136 case OP_TYPEMINUPTO:
2137 case OP_TYPEEXACT:
2138 case OP_TYPEPOSUPTO:
2139 if (code[1 + IMM2_SIZE] == OP_PROP || code[1 + IMM2_SIZE] == OP_NOTPROP)
2140 code += 2;
2141 break;
2142
2143 case OP_MARK:
2144 case OP_PRUNE_ARG:
2145 case OP_SKIP_ARG:
2146 case OP_THEN_ARG:
2147 code += code[1];
2148 break;
2149 }
2150
2151 /* Add in the fixed length from the table */
2152
2153 code += PRIV(OP_lengths)[c];
2154
2155 /* In UTF-8 mode, opcodes that are followed by a character may be followed by
2156 a multi-byte character. The length in the table is a minimum, so we have to
2157 arrange to skip the extra bytes. */
2158
2159 #if defined SUPPORT_UTF && !defined COMPILE_PCRE32
2160 if (utf) switch(c)
2161 {
2162 case OP_CHAR:
2163 case OP_CHARI:
2164 case OP_NOT:
2165 case OP_NOTI:
2166 case OP_EXACT:
2167 case OP_EXACTI:
2168 case OP_NOTEXACT:
2169 case OP_NOTEXACTI:
2170 case OP_UPTO:
2171 case OP_UPTOI:
2172 case OP_NOTUPTO:
2173 case OP_NOTUPTOI:
2174 case OP_MINUPTO:
2175 case OP_MINUPTOI:
2176 case OP_NOTMINUPTO:
2177 case OP_NOTMINUPTOI:
2178 case OP_POSUPTO:
2179 case OP_POSUPTOI:
2180 case OP_NOTPOSUPTO:
2181 case OP_NOTPOSUPTOI:
2182 case OP_STAR:
2183 case OP_STARI:
2184 case OP_NOTSTAR:
2185 case OP_NOTSTARI:
2186 case OP_MINSTAR:
2187 case OP_MINSTARI:
2188 case OP_NOTMINSTAR:
2189 case OP_NOTMINSTARI:
2190 case OP_POSSTAR:
2191 case OP_POSSTARI:
2192 case OP_NOTPOSSTAR:
2193 case OP_NOTPOSSTARI:
2194 case OP_PLUS:
2195 case OP_PLUSI:
2196 case OP_NOTPLUS:
2197 case OP_NOTPLUSI:
2198 case OP_MINPLUS:
2199 case OP_MINPLUSI:
2200 case OP_NOTMINPLUS:
2201 case OP_NOTMINPLUSI:
2202 case OP_POSPLUS:
2203 case OP_POSPLUSI:
2204 case OP_NOTPOSPLUS:
2205 case OP_NOTPOSPLUSI:
2206 case OP_QUERY:
2207 case OP_QUERYI:
2208 case OP_NOTQUERY:
2209 case OP_NOTQUERYI:
2210 case OP_MINQUERY:
2211 case OP_MINQUERYI:
2212 case OP_NOTMINQUERY:
2213 case OP_NOTMINQUERYI:
2214 case OP_POSQUERY:
2215 case OP_POSQUERYI:
2216 case OP_NOTPOSQUERY:
2217 case OP_NOTPOSQUERYI:
2218 if (HAS_EXTRALEN(code[-1])) code += GET_EXTRALEN(code[-1]);
2219 break;
2220 }
2221 #else
2222 (void)(utf); /* Keep compiler happy by referencing function argument */
2223 #endif
2224 }
2225 }
2226 }
2227
2228
2229
2230 /*************************************************
2231 * Scan compiled regex for recursion reference *
2232 *************************************************/
2233
2234 /* This little function scans through a compiled pattern until it finds an
2235 instance of OP_RECURSE.
2236
2237 Arguments:
2238 code points to start of expression
2239 utf TRUE in UTF-8 / UTF-16 / UTF-32 mode
2240
2241 Returns: pointer to the opcode for OP_RECURSE, or NULL if not found
2242 */
2243
2244 static const pcre_uchar *
find_recurse(const pcre_uchar * code,BOOL utf)2245 find_recurse(const pcre_uchar *code, BOOL utf)
2246 {
2247 for (;;)
2248 {
2249 register pcre_uchar c = *code;
2250 if (c == OP_END) return NULL;
2251 if (c == OP_RECURSE) return code;
2252
2253 /* XCLASS is used for classes that cannot be represented just by a bit
2254 map. This includes negated single high-valued characters. The length in
2255 the table is zero; the actual length is stored in the compiled code. */
2256
2257 if (c == OP_XCLASS) code += GET(code, 1);
2258
2259 /* Otherwise, we can get the item's length from the table, except that for
2260 repeated character types, we have to test for \p and \P, which have an extra
2261 two bytes of parameters, and for MARK/PRUNE/SKIP/THEN with an argument, we
2262 must add in its length. */
2263
2264 else
2265 {
2266 switch(c)
2267 {
2268 case OP_TYPESTAR:
2269 case OP_TYPEMINSTAR:
2270 case OP_TYPEPLUS:
2271 case OP_TYPEMINPLUS:
2272 case OP_TYPEQUERY:
2273 case OP_TYPEMINQUERY:
2274 case OP_TYPEPOSSTAR:
2275 case OP_TYPEPOSPLUS:
2276 case OP_TYPEPOSQUERY:
2277 if (code[1] == OP_PROP || code[1] == OP_NOTPROP) code += 2;
2278 break;
2279
2280 case OP_TYPEPOSUPTO:
2281 case OP_TYPEUPTO:
2282 case OP_TYPEMINUPTO:
2283 case OP_TYPEEXACT:
2284 if (code[1 + IMM2_SIZE] == OP_PROP || code[1 + IMM2_SIZE] == OP_NOTPROP)
2285 code += 2;
2286 break;
2287
2288 case OP_MARK:
2289 case OP_PRUNE_ARG:
2290 case OP_SKIP_ARG:
2291 case OP_THEN_ARG:
2292 code += code[1];
2293 break;
2294 }
2295
2296 /* Add in the fixed length from the table */
2297
2298 code += PRIV(OP_lengths)[c];
2299
2300 /* In UTF-8 mode, opcodes that are followed by a character may be followed
2301 by a multi-byte character. The length in the table is a minimum, so we have
2302 to arrange to skip the extra bytes. */
2303
2304 #if defined SUPPORT_UTF && !defined COMPILE_PCRE32
2305 if (utf) switch(c)
2306 {
2307 case OP_CHAR:
2308 case OP_CHARI:
2309 case OP_NOT:
2310 case OP_NOTI:
2311 case OP_EXACT:
2312 case OP_EXACTI:
2313 case OP_NOTEXACT:
2314 case OP_NOTEXACTI:
2315 case OP_UPTO:
2316 case OP_UPTOI:
2317 case OP_NOTUPTO:
2318 case OP_NOTUPTOI:
2319 case OP_MINUPTO:
2320 case OP_MINUPTOI:
2321 case OP_NOTMINUPTO:
2322 case OP_NOTMINUPTOI:
2323 case OP_POSUPTO:
2324 case OP_POSUPTOI:
2325 case OP_NOTPOSUPTO:
2326 case OP_NOTPOSUPTOI:
2327 case OP_STAR:
2328 case OP_STARI:
2329 case OP_NOTSTAR:
2330 case OP_NOTSTARI:
2331 case OP_MINSTAR:
2332 case OP_MINSTARI:
2333 case OP_NOTMINSTAR:
2334 case OP_NOTMINSTARI:
2335 case OP_POSSTAR:
2336 case OP_POSSTARI:
2337 case OP_NOTPOSSTAR:
2338 case OP_NOTPOSSTARI:
2339 case OP_PLUS:
2340 case OP_PLUSI:
2341 case OP_NOTPLUS:
2342 case OP_NOTPLUSI:
2343 case OP_MINPLUS:
2344 case OP_MINPLUSI:
2345 case OP_NOTMINPLUS:
2346 case OP_NOTMINPLUSI:
2347 case OP_POSPLUS:
2348 case OP_POSPLUSI:
2349 case OP_NOTPOSPLUS:
2350 case OP_NOTPOSPLUSI:
2351 case OP_QUERY:
2352 case OP_QUERYI:
2353 case OP_NOTQUERY:
2354 case OP_NOTQUERYI:
2355 case OP_MINQUERY:
2356 case OP_MINQUERYI:
2357 case OP_NOTMINQUERY:
2358 case OP_NOTMINQUERYI:
2359 case OP_POSQUERY:
2360 case OP_POSQUERYI:
2361 case OP_NOTPOSQUERY:
2362 case OP_NOTPOSQUERYI:
2363 if (HAS_EXTRALEN(code[-1])) code += GET_EXTRALEN(code[-1]);
2364 break;
2365 }
2366 #else
2367 (void)(utf); /* Keep compiler happy by referencing function argument */
2368 #endif
2369 }
2370 }
2371 }
2372
2373
2374
2375 /*************************************************
2376 * Scan compiled branch for non-emptiness *
2377 *************************************************/
2378
2379 /* This function scans through a branch of a compiled pattern to see whether it
2380 can match the empty string or not. It is called from could_be_empty()
2381 below and from compile_branch() when checking for an unlimited repeat of a
2382 group that can match nothing. Note that first_significant_code() skips over
2383 backward and negative forward assertions when its final argument is TRUE. If we
2384 hit an unclosed bracket, we return "empty" - this means we've struck an inner
2385 bracket whose current branch will already have been scanned.
2386
2387 Arguments:
2388 code points to start of search
2389 endcode points to where to stop
2390 utf TRUE if in UTF-8 / UTF-16 / UTF-32 mode
2391 cd contains pointers to tables etc.
2392 recurses chain of recurse_check to catch mutual recursion
2393
2394 Returns: TRUE if what is matched could be empty
2395 */
2396
2397 static BOOL
could_be_empty_branch(const pcre_uchar * code,const pcre_uchar * endcode,BOOL utf,compile_data * cd,recurse_check * recurses)2398 could_be_empty_branch(const pcre_uchar *code, const pcre_uchar *endcode,
2399 BOOL utf, compile_data *cd, recurse_check *recurses)
2400 {
2401 register pcre_uchar c;
2402 recurse_check this_recurse;
2403
2404 for (code = first_significant_code(code + PRIV(OP_lengths)[*code], TRUE);
2405 code < endcode;
2406 code = first_significant_code(code + PRIV(OP_lengths)[c], TRUE))
2407 {
2408 const pcre_uchar *ccode;
2409
2410 c = *code;
2411
2412 /* Skip over forward assertions; the other assertions are skipped by
2413 first_significant_code() with a TRUE final argument. */
2414
2415 if (c == OP_ASSERT)
2416 {
2417 do code += GET(code, 1); while (*code == OP_ALT);
2418 c = *code;
2419 continue;
2420 }
2421
2422 /* For a recursion/subroutine call, if its end has been reached, which
2423 implies a backward reference subroutine call, we can scan it. If it's a
2424 forward reference subroutine call, we can't. To detect forward reference
2425 we have to scan up the list that is kept in the workspace. This function is
2426 called only when doing the real compile, not during the pre-compile that
2427 measures the size of the compiled pattern. */
2428
2429 if (c == OP_RECURSE)
2430 {
2431 const pcre_uchar *scode = cd->start_code + GET(code, 1);
2432 const pcre_uchar *endgroup = scode;
2433 BOOL empty_branch;
2434
2435 /* Test for forward reference or uncompleted reference. This is disabled
2436 when called to scan a completed pattern by setting cd->start_workspace to
2437 NULL. */
2438
2439 if (cd->start_workspace != NULL)
2440 {
2441 const pcre_uchar *tcode;
2442 for (tcode = cd->start_workspace; tcode < cd->hwm; tcode += LINK_SIZE)
2443 if ((int)GET(tcode, 0) == (int)(code + 1 - cd->start_code)) return TRUE;
2444 if (GET(scode, 1) == 0) return TRUE; /* Unclosed */
2445 }
2446
2447 /* If the reference is to a completed group, we need to detect whether this
2448 is a recursive call, as otherwise there will be an infinite loop. If it is
2449 a recursion, just skip over it. Simple recursions are easily detected. For
2450 mutual recursions we keep a chain on the stack. */
2451
2452 do endgroup += GET(endgroup, 1); while (*endgroup == OP_ALT);
2453 if (code >= scode && code <= endgroup) continue; /* Simple recursion */
2454 else
2455 {
2456 recurse_check *r = recurses;
2457 for (r = recurses; r != NULL; r = r->prev)
2458 if (r->group == scode) break;
2459 if (r != NULL) continue; /* Mutual recursion */
2460 }
2461
2462 /* Completed reference; scan the referenced group, remembering it on the
2463 stack chain to detect mutual recursions. */
2464
2465 empty_branch = FALSE;
2466 this_recurse.prev = recurses;
2467 this_recurse.group = scode;
2468
2469 do
2470 {
2471 if (could_be_empty_branch(scode, endcode, utf, cd, &this_recurse))
2472 {
2473 empty_branch = TRUE;
2474 break;
2475 }
2476 scode += GET(scode, 1);
2477 }
2478 while (*scode == OP_ALT);
2479
2480 if (!empty_branch) return FALSE; /* All branches are non-empty */
2481 continue;
2482 }
2483
2484 /* Groups with zero repeats can of course be empty; skip them. */
2485
2486 if (c == OP_BRAZERO || c == OP_BRAMINZERO || c == OP_SKIPZERO ||
2487 c == OP_BRAPOSZERO)
2488 {
2489 code += PRIV(OP_lengths)[c];
2490 do code += GET(code, 1); while (*code == OP_ALT);
2491 c = *code;
2492 continue;
2493 }
2494
2495 /* A nested group that is already marked as "could be empty" can just be
2496 skipped. */
2497
2498 if (c == OP_SBRA || c == OP_SBRAPOS ||
2499 c == OP_SCBRA || c == OP_SCBRAPOS)
2500 {
2501 do code += GET(code, 1); while (*code == OP_ALT);
2502 c = *code;
2503 continue;
2504 }
2505
2506 /* For other groups, scan the branches. */
2507
2508 if (c == OP_BRA || c == OP_BRAPOS ||
2509 c == OP_CBRA || c == OP_CBRAPOS ||
2510 c == OP_ONCE || c == OP_ONCE_NC ||
2511 c == OP_COND || c == OP_SCOND)
2512 {
2513 BOOL empty_branch;
2514 if (GET(code, 1) == 0) return TRUE; /* Hit unclosed bracket */
2515
2516 /* If a conditional group has only one branch, there is a second, implied,
2517 empty branch, so just skip over the conditional, because it could be empty.
2518 Otherwise, scan the individual branches of the group. */
2519
2520 if (c == OP_COND && code[GET(code, 1)] != OP_ALT)
2521 code += GET(code, 1);
2522 else
2523 {
2524 empty_branch = FALSE;
2525 do
2526 {
2527 if (!empty_branch && could_be_empty_branch(code, endcode, utf, cd,
2528 recurses)) empty_branch = TRUE;
2529 code += GET(code, 1);
2530 }
2531 while (*code == OP_ALT);
2532 if (!empty_branch) return FALSE; /* All branches are non-empty */
2533 }
2534
2535 c = *code;
2536 continue;
2537 }
2538
2539 /* Handle the other opcodes */
2540
2541 switch (c)
2542 {
2543 /* Check for quantifiers after a class. XCLASS is used for classes that
2544 cannot be represented just by a bit map. This includes negated single
2545 high-valued characters. The length in PRIV(OP_lengths)[] is zero; the
2546 actual length is stored in the compiled code, so we must update "code"
2547 here. */
2548
2549 #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
2550 case OP_XCLASS:
2551 ccode = code += GET(code, 1);
2552 goto CHECK_CLASS_REPEAT;
2553 #endif
2554
2555 case OP_CLASS:
2556 case OP_NCLASS:
2557 ccode = code + PRIV(OP_lengths)[OP_CLASS];
2558
2559 #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
2560 CHECK_CLASS_REPEAT:
2561 #endif
2562
2563 switch (*ccode)
2564 {
2565 case OP_CRSTAR: /* These could be empty; continue */
2566 case OP_CRMINSTAR:
2567 case OP_CRQUERY:
2568 case OP_CRMINQUERY:
2569 case OP_CRPOSSTAR:
2570 case OP_CRPOSQUERY:
2571 break;
2572
2573 default: /* Non-repeat => class must match */
2574 case OP_CRPLUS: /* These repeats aren't empty */
2575 case OP_CRMINPLUS:
2576 case OP_CRPOSPLUS:
2577 return FALSE;
2578
2579 case OP_CRRANGE:
2580 case OP_CRMINRANGE:
2581 case OP_CRPOSRANGE:
2582 if (GET2(ccode, 1) > 0) return FALSE; /* Minimum > 0 */
2583 break;
2584 }
2585 break;
2586
2587 /* Opcodes that must match a character */
2588
2589 case OP_ANY:
2590 case OP_ALLANY:
2591 case OP_ANYBYTE:
2592
2593 case OP_PROP:
2594 case OP_NOTPROP:
2595 case OP_ANYNL:
2596
2597 case OP_NOT_HSPACE:
2598 case OP_HSPACE:
2599 case OP_NOT_VSPACE:
2600 case OP_VSPACE:
2601 case OP_EXTUNI:
2602
2603 case OP_NOT_DIGIT:
2604 case OP_DIGIT:
2605 case OP_NOT_WHITESPACE:
2606 case OP_WHITESPACE:
2607 case OP_NOT_WORDCHAR:
2608 case OP_WORDCHAR:
2609
2610 case OP_CHAR:
2611 case OP_CHARI:
2612 case OP_NOT:
2613 case OP_NOTI:
2614
2615 case OP_PLUS:
2616 case OP_PLUSI:
2617 case OP_MINPLUS:
2618 case OP_MINPLUSI:
2619
2620 case OP_NOTPLUS:
2621 case OP_NOTPLUSI:
2622 case OP_NOTMINPLUS:
2623 case OP_NOTMINPLUSI:
2624
2625 case OP_POSPLUS:
2626 case OP_POSPLUSI:
2627 case OP_NOTPOSPLUS:
2628 case OP_NOTPOSPLUSI:
2629
2630 case OP_EXACT:
2631 case OP_EXACTI:
2632 case OP_NOTEXACT:
2633 case OP_NOTEXACTI:
2634
2635 case OP_TYPEPLUS:
2636 case OP_TYPEMINPLUS:
2637 case OP_TYPEPOSPLUS:
2638 case OP_TYPEEXACT:
2639
2640 return FALSE;
2641
2642 /* These are going to continue, as they may be empty, but we have to
2643 fudge the length for the \p and \P cases. */
2644
2645 case OP_TYPESTAR:
2646 case OP_TYPEMINSTAR:
2647 case OP_TYPEPOSSTAR:
2648 case OP_TYPEQUERY:
2649 case OP_TYPEMINQUERY:
2650 case OP_TYPEPOSQUERY:
2651 if (code[1] == OP_PROP || code[1] == OP_NOTPROP) code += 2;
2652 break;
2653
2654 /* Same for these */
2655
2656 case OP_TYPEUPTO:
2657 case OP_TYPEMINUPTO:
2658 case OP_TYPEPOSUPTO:
2659 if (code[1 + IMM2_SIZE] == OP_PROP || code[1 + IMM2_SIZE] == OP_NOTPROP)
2660 code += 2;
2661 break;
2662
2663 /* End of branch */
2664
2665 case OP_KET:
2666 case OP_KETRMAX:
2667 case OP_KETRMIN:
2668 case OP_KETRPOS:
2669 case OP_ALT:
2670 return TRUE;
2671
2672 /* In UTF-8 mode, STAR, MINSTAR, POSSTAR, QUERY, MINQUERY, POSQUERY, UPTO,
2673 MINUPTO, and POSUPTO and their caseless and negative versions may be
2674 followed by a multibyte character. */
2675
2676 #if defined SUPPORT_UTF && !defined COMPILE_PCRE32
2677 case OP_STAR:
2678 case OP_STARI:
2679 case OP_NOTSTAR:
2680 case OP_NOTSTARI:
2681
2682 case OP_MINSTAR:
2683 case OP_MINSTARI:
2684 case OP_NOTMINSTAR:
2685 case OP_NOTMINSTARI:
2686
2687 case OP_POSSTAR:
2688 case OP_POSSTARI:
2689 case OP_NOTPOSSTAR:
2690 case OP_NOTPOSSTARI:
2691
2692 case OP_QUERY:
2693 case OP_QUERYI:
2694 case OP_NOTQUERY:
2695 case OP_NOTQUERYI:
2696
2697 case OP_MINQUERY:
2698 case OP_MINQUERYI:
2699 case OP_NOTMINQUERY:
2700 case OP_NOTMINQUERYI:
2701
2702 case OP_POSQUERY:
2703 case OP_POSQUERYI:
2704 case OP_NOTPOSQUERY:
2705 case OP_NOTPOSQUERYI:
2706
2707 if (utf && HAS_EXTRALEN(code[1])) code += GET_EXTRALEN(code[1]);
2708 break;
2709
2710 case OP_UPTO:
2711 case OP_UPTOI:
2712 case OP_NOTUPTO:
2713 case OP_NOTUPTOI:
2714
2715 case OP_MINUPTO:
2716 case OP_MINUPTOI:
2717 case OP_NOTMINUPTO:
2718 case OP_NOTMINUPTOI:
2719
2720 case OP_POSUPTO:
2721 case OP_POSUPTOI:
2722 case OP_NOTPOSUPTO:
2723 case OP_NOTPOSUPTOI:
2724
2725 if (utf && HAS_EXTRALEN(code[1 + IMM2_SIZE])) code += GET_EXTRALEN(code[1 + IMM2_SIZE]);
2726 break;
2727 #endif
2728
2729 /* MARK, and PRUNE/SKIP/THEN with an argument must skip over the argument
2730 string. */
2731
2732 case OP_MARK:
2733 case OP_PRUNE_ARG:
2734 case OP_SKIP_ARG:
2735 case OP_THEN_ARG:
2736 code += code[1];
2737 break;
2738
2739 /* None of the remaining opcodes are required to match a character. */
2740
2741 default:
2742 break;
2743 }
2744 }
2745
2746 return TRUE;
2747 }
2748
2749
2750
2751 /*************************************************
2752 * Scan compiled regex for non-emptiness *
2753 *************************************************/
2754
2755 /* This function is called to check for left recursive calls. We want to check
2756 the current branch of the current pattern to see if it could match the empty
2757 string. If it could, we must look outwards for branches at other levels,
2758 stopping when we pass beyond the bracket which is the subject of the recursion.
2759 This function is called only during the real compile, not during the
2760 pre-compile.
2761
2762 Arguments:
2763 code points to start of the recursion
2764 endcode points to where to stop (current RECURSE item)
2765 bcptr points to the chain of current (unclosed) branch starts
2766 utf TRUE if in UTF-8 / UTF-16 / UTF-32 mode
2767 cd pointers to tables etc
2768
2769 Returns: TRUE if what is matched could be empty
2770 */
2771
2772 static BOOL
could_be_empty(const pcre_uchar * code,const pcre_uchar * endcode,branch_chain * bcptr,BOOL utf,compile_data * cd)2773 could_be_empty(const pcre_uchar *code, const pcre_uchar *endcode,
2774 branch_chain *bcptr, BOOL utf, compile_data *cd)
2775 {
2776 while (bcptr != NULL && bcptr->current_branch >= code)
2777 {
2778 if (!could_be_empty_branch(bcptr->current_branch, endcode, utf, cd, NULL))
2779 return FALSE;
2780 bcptr = bcptr->outer;
2781 }
2782 return TRUE;
2783 }
2784
2785
2786
2787 /*************************************************
2788 * Base opcode of repeated opcodes *
2789 *************************************************/
2790
2791 /* Returns the base opcode for repeated single character type opcodes. If the
2792 opcode is not a repeated character type, it returns with the original value.
2793
2794 Arguments: c opcode
2795 Returns: base opcode for the type
2796 */
2797
2798 static pcre_uchar
get_repeat_base(pcre_uchar c)2799 get_repeat_base(pcre_uchar c)
2800 {
2801 return (c > OP_TYPEPOSUPTO)? c :
2802 (c >= OP_TYPESTAR)? OP_TYPESTAR :
2803 (c >= OP_NOTSTARI)? OP_NOTSTARI :
2804 (c >= OP_NOTSTAR)? OP_NOTSTAR :
2805 (c >= OP_STARI)? OP_STARI :
2806 OP_STAR;
2807 }
2808
2809
2810
2811 #ifdef SUPPORT_UCP
2812 /*************************************************
2813 * Check a character and a property *
2814 *************************************************/
2815
2816 /* This function is called by check_auto_possessive() when a property item
2817 is adjacent to a fixed character.
2818
2819 Arguments:
2820 c the character
2821 ptype the property type
2822 pdata the data for the type
2823 negated TRUE if it's a negated property (\P or \p{^)
2824
2825 Returns: TRUE if auto-possessifying is OK
2826 */
2827
2828 static BOOL
check_char_prop(pcre_uint32 c,unsigned int ptype,unsigned int pdata,BOOL negated)2829 check_char_prop(pcre_uint32 c, unsigned int ptype, unsigned int pdata,
2830 BOOL negated)
2831 {
2832 const pcre_uint32 *p;
2833 const ucd_record *prop = GET_UCD(c);
2834
2835 switch(ptype)
2836 {
2837 case PT_LAMP:
2838 return (prop->chartype == ucp_Lu ||
2839 prop->chartype == ucp_Ll ||
2840 prop->chartype == ucp_Lt) == negated;
2841
2842 case PT_GC:
2843 return (pdata == PRIV(ucp_gentype)[prop->chartype]) == negated;
2844
2845 case PT_PC:
2846 return (pdata == prop->chartype) == negated;
2847
2848 case PT_SC:
2849 return (pdata == prop->script) == negated;
2850
2851 /* These are specials */
2852
2853 case PT_ALNUM:
2854 return (PRIV(ucp_gentype)[prop->chartype] == ucp_L ||
2855 PRIV(ucp_gentype)[prop->chartype] == ucp_N) == negated;
2856
2857 /* Perl space used to exclude VT, but from Perl 5.18 it is included, which
2858 means that Perl space and POSIX space are now identical. PCRE was changed
2859 at release 8.34. */
2860
2861 case PT_SPACE: /* Perl space */
2862 case PT_PXSPACE: /* POSIX space */
2863 switch(c)
2864 {
2865 HSPACE_CASES:
2866 VSPACE_CASES:
2867 return negated;
2868
2869 default:
2870 return (PRIV(ucp_gentype)[prop->chartype] == ucp_Z) == negated;
2871 }
2872 break; /* Control never reaches here */
2873
2874 case PT_WORD:
2875 return (PRIV(ucp_gentype)[prop->chartype] == ucp_L ||
2876 PRIV(ucp_gentype)[prop->chartype] == ucp_N ||
2877 c == CHAR_UNDERSCORE) == negated;
2878
2879 case PT_CLIST:
2880 p = PRIV(ucd_caseless_sets) + prop->caseset;
2881 for (;;)
2882 {
2883 if (c < *p) return !negated;
2884 if (c == *p++) return negated;
2885 }
2886 break; /* Control never reaches here */
2887 }
2888
2889 return FALSE;
2890 }
2891 #endif /* SUPPORT_UCP */
2892
2893
2894
2895 /*************************************************
2896 * Fill the character property list *
2897 *************************************************/
2898
2899 /* Checks whether the code points to an opcode that can take part in auto-
2900 possessification, and if so, fills a list with its properties.
2901
2902 Arguments:
2903 code points to start of expression
2904 utf TRUE if in UTF-8 / UTF-16 / UTF-32 mode
2905 fcc points to case-flipping table
2906 list points to output list
2907 list[0] will be filled with the opcode
2908 list[1] will be non-zero if this opcode
2909 can match an empty character string
2910 list[2..7] depends on the opcode
2911
2912 Returns: points to the start of the next opcode if *code is accepted
2913 NULL if *code is not accepted
2914 */
2915
2916 static const pcre_uchar *
get_chr_property_list(const pcre_uchar * code,BOOL utf,const pcre_uint8 * fcc,pcre_uint32 * list)2917 get_chr_property_list(const pcre_uchar *code, BOOL utf,
2918 const pcre_uint8 *fcc, pcre_uint32 *list)
2919 {
2920 pcre_uchar c = *code;
2921 pcre_uchar base;
2922 const pcre_uchar *end;
2923 pcre_uint32 chr;
2924
2925 #ifdef SUPPORT_UCP
2926 pcre_uint32 *clist_dest;
2927 const pcre_uint32 *clist_src;
2928 #else
2929 utf = utf; /* Suppress "unused parameter" compiler warning */
2930 #endif
2931
2932 list[0] = c;
2933 list[1] = FALSE;
2934 code++;
2935
2936 if (c >= OP_STAR && c <= OP_TYPEPOSUPTO)
2937 {
2938 base = get_repeat_base(c);
2939 c -= (base - OP_STAR);
2940
2941 if (c == OP_UPTO || c == OP_MINUPTO || c == OP_EXACT || c == OP_POSUPTO)
2942 code += IMM2_SIZE;
2943
2944 list[1] = (c != OP_PLUS && c != OP_MINPLUS && c != OP_EXACT && c != OP_POSPLUS);
2945
2946 switch(base)
2947 {
2948 case OP_STAR:
2949 list[0] = OP_CHAR;
2950 break;
2951
2952 case OP_STARI:
2953 list[0] = OP_CHARI;
2954 break;
2955
2956 case OP_NOTSTAR:
2957 list[0] = OP_NOT;
2958 break;
2959
2960 case OP_NOTSTARI:
2961 list[0] = OP_NOTI;
2962 break;
2963
2964 case OP_TYPESTAR:
2965 list[0] = *code;
2966 code++;
2967 break;
2968 }
2969 c = list[0];
2970 }
2971
2972 switch(c)
2973 {
2974 case OP_NOT_DIGIT:
2975 case OP_DIGIT:
2976 case OP_NOT_WHITESPACE:
2977 case OP_WHITESPACE:
2978 case OP_NOT_WORDCHAR:
2979 case OP_WORDCHAR:
2980 case OP_ANY:
2981 case OP_ALLANY:
2982 case OP_ANYNL:
2983 case OP_NOT_HSPACE:
2984 case OP_HSPACE:
2985 case OP_NOT_VSPACE:
2986 case OP_VSPACE:
2987 case OP_EXTUNI:
2988 case OP_EODN:
2989 case OP_EOD:
2990 case OP_DOLL:
2991 case OP_DOLLM:
2992 return code;
2993
2994 case OP_CHAR:
2995 case OP_NOT:
2996 GETCHARINCTEST(chr, code);
2997 list[2] = chr;
2998 list[3] = NOTACHAR;
2999 return code;
3000
3001 case OP_CHARI:
3002 case OP_NOTI:
3003 list[0] = (c == OP_CHARI) ? OP_CHAR : OP_NOT;
3004 GETCHARINCTEST(chr, code);
3005 list[2] = chr;
3006
3007 #ifdef SUPPORT_UCP
3008 if (chr < 128 || (chr < 256 && !utf))
3009 list[3] = fcc[chr];
3010 else
3011 list[3] = UCD_OTHERCASE(chr);
3012 #elif defined SUPPORT_UTF || !defined COMPILE_PCRE8
3013 list[3] = (chr < 256) ? fcc[chr] : chr;
3014 #else
3015 list[3] = fcc[chr];
3016 #endif
3017
3018 /* The othercase might be the same value. */
3019
3020 if (chr == list[3])
3021 list[3] = NOTACHAR;
3022 else
3023 list[4] = NOTACHAR;
3024 return code;
3025
3026 #ifdef SUPPORT_UCP
3027 case OP_PROP:
3028 case OP_NOTPROP:
3029 if (code[0] != PT_CLIST)
3030 {
3031 list[2] = code[0];
3032 list[3] = code[1];
3033 return code + 2;
3034 }
3035
3036 /* Convert only if we have enough space. */
3037
3038 clist_src = PRIV(ucd_caseless_sets) + code[1];
3039 clist_dest = list + 2;
3040 code += 2;
3041
3042 do {
3043 if (clist_dest >= list + 8)
3044 {
3045 /* Early return if there is not enough space. This should never
3046 happen, since all clists are shorter than 5 character now. */
3047 list[2] = code[0];
3048 list[3] = code[1];
3049 return code;
3050 }
3051 *clist_dest++ = *clist_src;
3052 }
3053 while(*clist_src++ != NOTACHAR);
3054
3055 /* All characters are stored. The terminating NOTACHAR
3056 is copied form the clist itself. */
3057
3058 list[0] = (c == OP_PROP) ? OP_CHAR : OP_NOT;
3059 return code;
3060 #endif
3061
3062 case OP_NCLASS:
3063 case OP_CLASS:
3064 #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
3065 case OP_XCLASS:
3066 if (c == OP_XCLASS)
3067 end = code + GET(code, 0) - 1;
3068 else
3069 #endif
3070 end = code + 32 / sizeof(pcre_uchar);
3071
3072 switch(*end)
3073 {
3074 case OP_CRSTAR:
3075 case OP_CRMINSTAR:
3076 case OP_CRQUERY:
3077 case OP_CRMINQUERY:
3078 case OP_CRPOSSTAR:
3079 case OP_CRPOSQUERY:
3080 list[1] = TRUE;
3081 end++;
3082 break;
3083
3084 case OP_CRPLUS:
3085 case OP_CRMINPLUS:
3086 case OP_CRPOSPLUS:
3087 end++;
3088 break;
3089
3090 case OP_CRRANGE:
3091 case OP_CRMINRANGE:
3092 case OP_CRPOSRANGE:
3093 list[1] = (GET2(end, 1) == 0);
3094 end += 1 + 2 * IMM2_SIZE;
3095 break;
3096 }
3097 list[2] = (pcre_uint32)(end - code);
3098 return end;
3099 }
3100 return NULL; /* Opcode not accepted */
3101 }
3102
3103
3104
3105 /*************************************************
3106 * Scan further character sets for match *
3107 *************************************************/
3108
3109 /* Checks whether the base and the current opcode have a common character, in
3110 which case the base cannot be possessified.
3111
3112 Arguments:
3113 code points to the byte code
3114 utf TRUE in UTF-8 / UTF-16 / UTF-32 mode
3115 cd static compile data
3116 base_list the data list of the base opcode
3117
3118 Returns: TRUE if the auto-possessification is possible
3119 */
3120
3121 static BOOL
compare_opcodes(const pcre_uchar * code,BOOL utf,const compile_data * cd,const pcre_uint32 * base_list,const pcre_uchar * base_end,int * rec_limit)3122 compare_opcodes(const pcre_uchar *code, BOOL utf, const compile_data *cd,
3123 const pcre_uint32 *base_list, const pcre_uchar *base_end, int *rec_limit)
3124 {
3125 pcre_uchar c;
3126 pcre_uint32 list[8];
3127 const pcre_uint32 *chr_ptr;
3128 const pcre_uint32 *ochr_ptr;
3129 const pcre_uint32 *list_ptr;
3130 const pcre_uchar *next_code;
3131 #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
3132 const pcre_uchar *xclass_flags;
3133 #endif
3134 const pcre_uint8 *class_bitset;
3135 const pcre_uint8 *set1, *set2, *set_end;
3136 pcre_uint32 chr;
3137 BOOL accepted, invert_bits;
3138 BOOL entered_a_group = FALSE;
3139
3140 if (*rec_limit == 0) return FALSE;
3141 --(*rec_limit);
3142
3143 /* Note: the base_list[1] contains whether the current opcode has greedy
3144 (represented by a non-zero value) quantifier. This is a different from
3145 other character type lists, which stores here that the character iterator
3146 matches to an empty string (also represented by a non-zero value). */
3147
3148 for(;;)
3149 {
3150 /* All operations move the code pointer forward.
3151 Therefore infinite recursions are not possible. */
3152
3153 c = *code;
3154
3155 /* Skip over callouts */
3156
3157 if (c == OP_CALLOUT)
3158 {
3159 code += PRIV(OP_lengths)[c];
3160 continue;
3161 }
3162
3163 if (c == OP_ALT)
3164 {
3165 do code += GET(code, 1); while (*code == OP_ALT);
3166 c = *code;
3167 }
3168
3169 switch(c)
3170 {
3171 case OP_END:
3172 case OP_KETRPOS:
3173 /* TRUE only in greedy case. The non-greedy case could be replaced by
3174 an OP_EXACT, but it is probably not worth it. (And note that OP_EXACT
3175 uses more memory, which we cannot get at this stage.) */
3176
3177 return base_list[1] != 0;
3178
3179 case OP_KET:
3180 /* If the bracket is capturing, and referenced by an OP_RECURSE, or
3181 it is an atomic sub-pattern (assert, once, etc.) the non-greedy case
3182 cannot be converted to a possessive form. */
3183
3184 if (base_list[1] == 0) return FALSE;
3185
3186 switch(*(code - GET(code, 1)))
3187 {
3188 case OP_ASSERT:
3189 case OP_ASSERT_NOT:
3190 case OP_ASSERTBACK:
3191 case OP_ASSERTBACK_NOT:
3192 case OP_ONCE:
3193 case OP_ONCE_NC:
3194 /* Atomic sub-patterns and assertions can always auto-possessify their
3195 last iterator. However, if the group was entered as a result of checking
3196 a previous iterator, this is not possible. */
3197
3198 return !entered_a_group;
3199 }
3200
3201 code += PRIV(OP_lengths)[c];
3202 continue;
3203
3204 case OP_ONCE:
3205 case OP_ONCE_NC:
3206 case OP_BRA:
3207 case OP_CBRA:
3208 next_code = code + GET(code, 1);
3209 code += PRIV(OP_lengths)[c];
3210
3211 while (*next_code == OP_ALT)
3212 {
3213 if (!compare_opcodes(code, utf, cd, base_list, base_end, rec_limit))
3214 return FALSE;
3215 code = next_code + 1 + LINK_SIZE;
3216 next_code += GET(next_code, 1);
3217 }
3218
3219 entered_a_group = TRUE;
3220 continue;
3221
3222 case OP_BRAZERO:
3223 case OP_BRAMINZERO:
3224
3225 next_code = code + 1;
3226 if (*next_code != OP_BRA && *next_code != OP_CBRA
3227 && *next_code != OP_ONCE && *next_code != OP_ONCE_NC) return FALSE;
3228
3229 do next_code += GET(next_code, 1); while (*next_code == OP_ALT);
3230
3231 /* The bracket content will be checked by the
3232 OP_BRA/OP_CBRA case above. */
3233 next_code += 1 + LINK_SIZE;
3234 if (!compare_opcodes(next_code, utf, cd, base_list, base_end, rec_limit))
3235 return FALSE;
3236
3237 code += PRIV(OP_lengths)[c];
3238 continue;
3239
3240 default:
3241 break;
3242 }
3243
3244 /* Check for a supported opcode, and load its properties. */
3245
3246 code = get_chr_property_list(code, utf, cd->fcc, list);
3247 if (code == NULL) return FALSE; /* Unsupported */
3248
3249 /* If either opcode is a small character list, set pointers for comparing
3250 characters from that list with another list, or with a property. */
3251
3252 if (base_list[0] == OP_CHAR)
3253 {
3254 chr_ptr = base_list + 2;
3255 list_ptr = list;
3256 }
3257 else if (list[0] == OP_CHAR)
3258 {
3259 chr_ptr = list + 2;
3260 list_ptr = base_list;
3261 }
3262
3263 /* Character bitsets can also be compared to certain opcodes. */
3264
3265 else if (base_list[0] == OP_CLASS || list[0] == OP_CLASS
3266 #ifdef COMPILE_PCRE8
3267 /* In 8 bit, non-UTF mode, OP_CLASS and OP_NCLASS are the same. */
3268 || (!utf && (base_list[0] == OP_NCLASS || list[0] == OP_NCLASS))
3269 #endif
3270 )
3271 {
3272 #ifdef COMPILE_PCRE8
3273 if (base_list[0] == OP_CLASS || (!utf && base_list[0] == OP_NCLASS))
3274 #else
3275 if (base_list[0] == OP_CLASS)
3276 #endif
3277 {
3278 set1 = (pcre_uint8 *)(base_end - base_list[2]);
3279 list_ptr = list;
3280 }
3281 else
3282 {
3283 set1 = (pcre_uint8 *)(code - list[2]);
3284 list_ptr = base_list;
3285 }
3286
3287 invert_bits = FALSE;
3288 switch(list_ptr[0])
3289 {
3290 case OP_CLASS:
3291 case OP_NCLASS:
3292 set2 = (pcre_uint8 *)
3293 ((list_ptr == list ? code : base_end) - list_ptr[2]);
3294 break;
3295
3296 #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
3297 case OP_XCLASS:
3298 xclass_flags = (list_ptr == list ? code : base_end) - list_ptr[2] + LINK_SIZE;
3299 if ((*xclass_flags & XCL_HASPROP) != 0) return FALSE;
3300 if ((*xclass_flags & XCL_MAP) == 0)
3301 {
3302 /* No bits are set for characters < 256. */
3303 if (list[1] == 0) return (*xclass_flags & XCL_NOT) == 0;
3304 /* Might be an empty repeat. */
3305 continue;
3306 }
3307 set2 = (pcre_uint8 *)(xclass_flags + 1);
3308 break;
3309 #endif
3310
3311 case OP_NOT_DIGIT:
3312 invert_bits = TRUE;
3313 /* Fall through */
3314 case OP_DIGIT:
3315 set2 = (pcre_uint8 *)(cd->cbits + cbit_digit);
3316 break;
3317
3318 case OP_NOT_WHITESPACE:
3319 invert_bits = TRUE;
3320 /* Fall through */
3321 case OP_WHITESPACE:
3322 set2 = (pcre_uint8 *)(cd->cbits + cbit_space);
3323 break;
3324
3325 case OP_NOT_WORDCHAR:
3326 invert_bits = TRUE;
3327 /* Fall through */
3328 case OP_WORDCHAR:
3329 set2 = (pcre_uint8 *)(cd->cbits + cbit_word);
3330 break;
3331
3332 default:
3333 return FALSE;
3334 }
3335
3336 /* Because the sets are unaligned, we need
3337 to perform byte comparison here. */
3338 set_end = set1 + 32;
3339 if (invert_bits)
3340 {
3341 do
3342 {
3343 if ((*set1++ & ~(*set2++)) != 0) return FALSE;
3344 }
3345 while (set1 < set_end);
3346 }
3347 else
3348 {
3349 do
3350 {
3351 if ((*set1++ & *set2++) != 0) return FALSE;
3352 }
3353 while (set1 < set_end);
3354 }
3355
3356 if (list[1] == 0) return TRUE;
3357 /* Might be an empty repeat. */
3358 continue;
3359 }
3360
3361 /* Some property combinations also acceptable. Unicode property opcodes are
3362 processed specially; the rest can be handled with a lookup table. */
3363
3364 else
3365 {
3366 pcre_uint32 leftop, rightop;
3367
3368 leftop = base_list[0];
3369 rightop = list[0];
3370
3371 #ifdef SUPPORT_UCP
3372 accepted = FALSE; /* Always set in non-unicode case. */
3373 if (leftop == OP_PROP || leftop == OP_NOTPROP)
3374 {
3375 if (rightop == OP_EOD)
3376 accepted = TRUE;
3377 else if (rightop == OP_PROP || rightop == OP_NOTPROP)
3378 {
3379 int n;
3380 const pcre_uint8 *p;
3381 BOOL same = leftop == rightop;
3382 BOOL lisprop = leftop == OP_PROP;
3383 BOOL risprop = rightop == OP_PROP;
3384 BOOL bothprop = lisprop && risprop;
3385
3386 /* There's a table that specifies how each combination is to be
3387 processed:
3388 0 Always return FALSE (never auto-possessify)
3389 1 Character groups are distinct (possessify if both are OP_PROP)
3390 2 Check character categories in the same group (general or particular)
3391 3 Return TRUE if the two opcodes are not the same
3392 ... see comments below
3393 */
3394
3395 n = propposstab[base_list[2]][list[2]];
3396 switch(n)
3397 {
3398 case 0: break;
3399 case 1: accepted = bothprop; break;
3400 case 2: accepted = (base_list[3] == list[3]) != same; break;
3401 case 3: accepted = !same; break;
3402
3403 case 4: /* Left general category, right particular category */
3404 accepted = risprop && catposstab[base_list[3]][list[3]] == same;
3405 break;
3406
3407 case 5: /* Right general category, left particular category */
3408 accepted = lisprop && catposstab[list[3]][base_list[3]] == same;
3409 break;
3410
3411 /* This code is logically tricky. Think hard before fiddling with it.
3412 The posspropstab table has four entries per row. Each row relates to
3413 one of PCRE's special properties such as ALNUM or SPACE or WORD.
3414 Only WORD actually needs all four entries, but using repeats for the
3415 others means they can all use the same code below.
3416
3417 The first two entries in each row are Unicode general categories, and
3418 apply always, because all the characters they include are part of the
3419 PCRE character set. The third and fourth entries are a general and a
3420 particular category, respectively, that include one or more relevant
3421 characters. One or the other is used, depending on whether the check
3422 is for a general or a particular category. However, in both cases the
3423 category contains more characters than the specials that are defined
3424 for the property being tested against. Therefore, it cannot be used
3425 in a NOTPROP case.
3426
3427 Example: the row for WORD contains ucp_L, ucp_N, ucp_P, ucp_Po.
3428 Underscore is covered by ucp_P or ucp_Po. */
3429
3430 case 6: /* Left alphanum vs right general category */
3431 case 7: /* Left space vs right general category */
3432 case 8: /* Left word vs right general category */
3433 p = posspropstab[n-6];
3434 accepted = risprop && lisprop ==
3435 (list[3] != p[0] &&
3436 list[3] != p[1] &&
3437 (list[3] != p[2] || !lisprop));
3438 break;
3439
3440 case 9: /* Right alphanum vs left general category */
3441 case 10: /* Right space vs left general category */
3442 case 11: /* Right word vs left general category */
3443 p = posspropstab[n-9];
3444 accepted = lisprop && risprop ==
3445 (base_list[3] != p[0] &&
3446 base_list[3] != p[1] &&
3447 (base_list[3] != p[2] || !risprop));
3448 break;
3449
3450 case 12: /* Left alphanum vs right particular category */
3451 case 13: /* Left space vs right particular category */
3452 case 14: /* Left word vs right particular category */
3453 p = posspropstab[n-12];
3454 accepted = risprop && lisprop ==
3455 (catposstab[p[0]][list[3]] &&
3456 catposstab[p[1]][list[3]] &&
3457 (list[3] != p[3] || !lisprop));
3458 break;
3459
3460 case 15: /* Right alphanum vs left particular category */
3461 case 16: /* Right space vs left particular category */
3462 case 17: /* Right word vs left particular category */
3463 p = posspropstab[n-15];
3464 accepted = lisprop && risprop ==
3465 (catposstab[p[0]][base_list[3]] &&
3466 catposstab[p[1]][base_list[3]] &&
3467 (base_list[3] != p[3] || !risprop));
3468 break;
3469 }
3470 }
3471 }
3472
3473 else
3474 #endif /* SUPPORT_UCP */
3475
3476 accepted = leftop >= FIRST_AUTOTAB_OP && leftop <= LAST_AUTOTAB_LEFT_OP &&
3477 rightop >= FIRST_AUTOTAB_OP && rightop <= LAST_AUTOTAB_RIGHT_OP &&
3478 autoposstab[leftop - FIRST_AUTOTAB_OP][rightop - FIRST_AUTOTAB_OP];
3479
3480 if (!accepted) return FALSE;
3481
3482 if (list[1] == 0) return TRUE;
3483 /* Might be an empty repeat. */
3484 continue;
3485 }
3486
3487 /* Control reaches here only if one of the items is a small character list.
3488 All characters are checked against the other side. */
3489
3490 do
3491 {
3492 chr = *chr_ptr;
3493
3494 switch(list_ptr[0])
3495 {
3496 case OP_CHAR:
3497 ochr_ptr = list_ptr + 2;
3498 do
3499 {
3500 if (chr == *ochr_ptr) return FALSE;
3501 ochr_ptr++;
3502 }
3503 while(*ochr_ptr != NOTACHAR);
3504 break;
3505
3506 case OP_NOT:
3507 ochr_ptr = list_ptr + 2;
3508 do
3509 {
3510 if (chr == *ochr_ptr)
3511 break;
3512 ochr_ptr++;
3513 }
3514 while(*ochr_ptr != NOTACHAR);
3515 if (*ochr_ptr == NOTACHAR) return FALSE; /* Not found */
3516 break;
3517
3518 /* Note that OP_DIGIT etc. are generated only when PCRE_UCP is *not*
3519 set. When it is set, \d etc. are converted into OP_(NOT_)PROP codes. */
3520
3521 case OP_DIGIT:
3522 if (chr < 256 && (cd->ctypes[chr] & ctype_digit) != 0) return FALSE;
3523 break;
3524
3525 case OP_NOT_DIGIT:
3526 if (chr > 255 || (cd->ctypes[chr] & ctype_digit) == 0) return FALSE;
3527 break;
3528
3529 case OP_WHITESPACE:
3530 if (chr < 256 && (cd->ctypes[chr] & ctype_space) != 0) return FALSE;
3531 break;
3532
3533 case OP_NOT_WHITESPACE:
3534 if (chr > 255 || (cd->ctypes[chr] & ctype_space) == 0) return FALSE;
3535 break;
3536
3537 case OP_WORDCHAR:
3538 if (chr < 255 && (cd->ctypes[chr] & ctype_word) != 0) return FALSE;
3539 break;
3540
3541 case OP_NOT_WORDCHAR:
3542 if (chr > 255 || (cd->ctypes[chr] & ctype_word) == 0) return FALSE;
3543 break;
3544
3545 case OP_HSPACE:
3546 switch(chr)
3547 {
3548 HSPACE_CASES: return FALSE;
3549 default: break;
3550 }
3551 break;
3552
3553 case OP_NOT_HSPACE:
3554 switch(chr)
3555 {
3556 HSPACE_CASES: break;
3557 default: return FALSE;
3558 }
3559 break;
3560
3561 case OP_ANYNL:
3562 case OP_VSPACE:
3563 switch(chr)
3564 {
3565 VSPACE_CASES: return FALSE;
3566 default: break;
3567 }
3568 break;
3569
3570 case OP_NOT_VSPACE:
3571 switch(chr)
3572 {
3573 VSPACE_CASES: break;
3574 default: return FALSE;
3575 }
3576 break;
3577
3578 case OP_DOLL:
3579 case OP_EODN:
3580 switch (chr)
3581 {
3582 case CHAR_CR:
3583 case CHAR_LF:
3584 case CHAR_VT:
3585 case CHAR_FF:
3586 case CHAR_NEL:
3587 #ifndef EBCDIC
3588 case 0x2028:
3589 case 0x2029:
3590 #endif /* Not EBCDIC */
3591 return FALSE;
3592 }
3593 break;
3594
3595 case OP_EOD: /* Can always possessify before \z */
3596 break;
3597
3598 #ifdef SUPPORT_UCP
3599 case OP_PROP:
3600 case OP_NOTPROP:
3601 if (!check_char_prop(chr, list_ptr[2], list_ptr[3],
3602 list_ptr[0] == OP_NOTPROP))
3603 return FALSE;
3604 break;
3605 #endif
3606
3607 case OP_NCLASS:
3608 if (chr > 255) return FALSE;
3609 /* Fall through */
3610
3611 case OP_CLASS:
3612 if (chr > 255) break;
3613 class_bitset = (pcre_uint8 *)
3614 ((list_ptr == list ? code : base_end) - list_ptr[2]);
3615 if ((class_bitset[chr >> 3] & (1U << (chr & 7))) != 0) return FALSE;
3616 break;
3617
3618 #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
3619 case OP_XCLASS:
3620 if (PRIV(xclass)(chr, (list_ptr == list ? code : base_end) -
3621 list_ptr[2] + LINK_SIZE, utf)) return FALSE;
3622 break;
3623 #endif
3624
3625 default:
3626 return FALSE;
3627 }
3628
3629 chr_ptr++;
3630 }
3631 while(*chr_ptr != NOTACHAR);
3632
3633 /* At least one character must be matched from this opcode. */
3634
3635 if (list[1] == 0) return TRUE;
3636 }
3637
3638 /* Control never reaches here. There used to be a fail-save return FALSE; here,
3639 but some compilers complain about an unreachable statement. */
3640
3641 }
3642
3643
3644
3645 /*************************************************
3646 * Scan compiled regex for auto-possession *
3647 *************************************************/
3648
3649 /* Replaces single character iterations with their possessive alternatives
3650 if appropriate. This function modifies the compiled opcode!
3651
3652 Arguments:
3653 code points to start of the byte code
3654 utf TRUE in UTF-8 / UTF-16 / UTF-32 mode
3655 cd static compile data
3656
3657 Returns: nothing
3658 */
3659
3660 static void
auto_possessify(pcre_uchar * code,BOOL utf,const compile_data * cd)3661 auto_possessify(pcre_uchar *code, BOOL utf, const compile_data *cd)
3662 {
3663 register pcre_uchar c;
3664 const pcre_uchar *end;
3665 pcre_uchar *repeat_opcode;
3666 pcre_uint32 list[8];
3667 int rec_limit;
3668
3669 for (;;)
3670 {
3671 c = *code;
3672
3673 /* When a pattern with bad UTF-8 encoding is compiled with NO_UTF_CHECK,
3674 it may compile without complaining, but may get into a loop here if the code
3675 pointer points to a bad value. This is, of course a documentated possibility,
3676 when NO_UTF_CHECK is set, so it isn't a bug, but we can detect this case and
3677 just give up on this optimization. */
3678
3679 if (c >= OP_TABLE_LENGTH) return;
3680
3681 if (c >= OP_STAR && c <= OP_TYPEPOSUPTO)
3682 {
3683 c -= get_repeat_base(c) - OP_STAR;
3684 end = (c <= OP_MINUPTO) ?
3685 get_chr_property_list(code, utf, cd->fcc, list) : NULL;
3686 list[1] = c == OP_STAR || c == OP_PLUS || c == OP_QUERY || c == OP_UPTO;
3687
3688 rec_limit = 1000;
3689 if (end != NULL && compare_opcodes(end, utf, cd, list, end, &rec_limit))
3690 {
3691 switch(c)
3692 {
3693 case OP_STAR:
3694 *code += OP_POSSTAR - OP_STAR;
3695 break;
3696
3697 case OP_MINSTAR:
3698 *code += OP_POSSTAR - OP_MINSTAR;
3699 break;
3700
3701 case OP_PLUS:
3702 *code += OP_POSPLUS - OP_PLUS;
3703 break;
3704
3705 case OP_MINPLUS:
3706 *code += OP_POSPLUS - OP_MINPLUS;
3707 break;
3708
3709 case OP_QUERY:
3710 *code += OP_POSQUERY - OP_QUERY;
3711 break;
3712
3713 case OP_MINQUERY:
3714 *code += OP_POSQUERY - OP_MINQUERY;
3715 break;
3716
3717 case OP_UPTO:
3718 *code += OP_POSUPTO - OP_UPTO;
3719 break;
3720
3721 case OP_MINUPTO:
3722 *code += OP_POSUPTO - OP_MINUPTO;
3723 break;
3724 }
3725 }
3726 c = *code;
3727 }
3728 else if (c == OP_CLASS || c == OP_NCLASS || c == OP_XCLASS)
3729 {
3730 #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
3731 if (c == OP_XCLASS)
3732 repeat_opcode = code + GET(code, 1);
3733 else
3734 #endif
3735 repeat_opcode = code + 1 + (32 / sizeof(pcre_uchar));
3736
3737 c = *repeat_opcode;
3738 if (c >= OP_CRSTAR && c <= OP_CRMINRANGE)
3739 {
3740 /* end must not be NULL. */
3741 end = get_chr_property_list(code, utf, cd->fcc, list);
3742
3743 list[1] = (c & 1) == 0;
3744
3745 rec_limit = 1000;
3746 if (compare_opcodes(end, utf, cd, list, end, &rec_limit))
3747 {
3748 switch (c)
3749 {
3750 case OP_CRSTAR:
3751 case OP_CRMINSTAR:
3752 *repeat_opcode = OP_CRPOSSTAR;
3753 break;
3754
3755 case OP_CRPLUS:
3756 case OP_CRMINPLUS:
3757 *repeat_opcode = OP_CRPOSPLUS;
3758 break;
3759
3760 case OP_CRQUERY:
3761 case OP_CRMINQUERY:
3762 *repeat_opcode = OP_CRPOSQUERY;
3763 break;
3764
3765 case OP_CRRANGE:
3766 case OP_CRMINRANGE:
3767 *repeat_opcode = OP_CRPOSRANGE;
3768 break;
3769 }
3770 }
3771 }
3772 c = *code;
3773 }
3774
3775 switch(c)
3776 {
3777 case OP_END:
3778 return;
3779
3780 case OP_TYPESTAR:
3781 case OP_TYPEMINSTAR:
3782 case OP_TYPEPLUS:
3783 case OP_TYPEMINPLUS:
3784 case OP_TYPEQUERY:
3785 case OP_TYPEMINQUERY:
3786 case OP_TYPEPOSSTAR:
3787 case OP_TYPEPOSPLUS:
3788 case OP_TYPEPOSQUERY:
3789 if (code[1] == OP_PROP || code[1] == OP_NOTPROP) code += 2;
3790 break;
3791
3792 case OP_TYPEUPTO:
3793 case OP_TYPEMINUPTO:
3794 case OP_TYPEEXACT:
3795 case OP_TYPEPOSUPTO:
3796 if (code[1 + IMM2_SIZE] == OP_PROP || code[1 + IMM2_SIZE] == OP_NOTPROP)
3797 code += 2;
3798 break;
3799
3800 #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
3801 case OP_XCLASS:
3802 code += GET(code, 1);
3803 break;
3804 #endif
3805
3806 case OP_MARK:
3807 case OP_PRUNE_ARG:
3808 case OP_SKIP_ARG:
3809 case OP_THEN_ARG:
3810 code += code[1];
3811 break;
3812 }
3813
3814 /* Add in the fixed length from the table */
3815
3816 code += PRIV(OP_lengths)[c];
3817
3818 /* In UTF-8 mode, opcodes that are followed by a character may be followed by
3819 a multi-byte character. The length in the table is a minimum, so we have to
3820 arrange to skip the extra bytes. */
3821
3822 #if defined SUPPORT_UTF && !defined COMPILE_PCRE32
3823 if (utf) switch(c)
3824 {
3825 case OP_CHAR:
3826 case OP_CHARI:
3827 case OP_NOT:
3828 case OP_NOTI:
3829 case OP_STAR:
3830 case OP_MINSTAR:
3831 case OP_PLUS:
3832 case OP_MINPLUS:
3833 case OP_QUERY:
3834 case OP_MINQUERY:
3835 case OP_UPTO:
3836 case OP_MINUPTO:
3837 case OP_EXACT:
3838 case OP_POSSTAR:
3839 case OP_POSPLUS:
3840 case OP_POSQUERY:
3841 case OP_POSUPTO:
3842 case OP_STARI:
3843 case OP_MINSTARI:
3844 case OP_PLUSI:
3845 case OP_MINPLUSI:
3846 case OP_QUERYI:
3847 case OP_MINQUERYI:
3848 case OP_UPTOI:
3849 case OP_MINUPTOI:
3850 case OP_EXACTI:
3851 case OP_POSSTARI:
3852 case OP_POSPLUSI:
3853 case OP_POSQUERYI:
3854 case OP_POSUPTOI:
3855 case OP_NOTSTAR:
3856 case OP_NOTMINSTAR:
3857 case OP_NOTPLUS:
3858 case OP_NOTMINPLUS:
3859 case OP_NOTQUERY:
3860 case OP_NOTMINQUERY:
3861 case OP_NOTUPTO:
3862 case OP_NOTMINUPTO:
3863 case OP_NOTEXACT:
3864 case OP_NOTPOSSTAR:
3865 case OP_NOTPOSPLUS:
3866 case OP_NOTPOSQUERY:
3867 case OP_NOTPOSUPTO:
3868 case OP_NOTSTARI:
3869 case OP_NOTMINSTARI:
3870 case OP_NOTPLUSI:
3871 case OP_NOTMINPLUSI:
3872 case OP_NOTQUERYI:
3873 case OP_NOTMINQUERYI:
3874 case OP_NOTUPTOI:
3875 case OP_NOTMINUPTOI:
3876 case OP_NOTEXACTI:
3877 case OP_NOTPOSSTARI:
3878 case OP_NOTPOSPLUSI:
3879 case OP_NOTPOSQUERYI:
3880 case OP_NOTPOSUPTOI:
3881 if (HAS_EXTRALEN(code[-1])) code += GET_EXTRALEN(code[-1]);
3882 break;
3883 }
3884 #else
3885 (void)(utf); /* Keep compiler happy by referencing function argument */
3886 #endif
3887 }
3888 }
3889
3890
3891
3892 /*************************************************
3893 * Check for POSIX class syntax *
3894 *************************************************/
3895
3896 /* This function is called when the sequence "[:" or "[." or "[=" is
3897 encountered in a character class. It checks whether this is followed by a
3898 sequence of characters terminated by a matching ":]" or ".]" or "=]". If we
3899 reach an unescaped ']' without the special preceding character, return FALSE.
3900
3901 Originally, this function only recognized a sequence of letters between the
3902 terminators, but it seems that Perl recognizes any sequence of characters,
3903 though of course unknown POSIX names are subsequently rejected. Perl gives an
3904 "Unknown POSIX class" error for [:f\oo:] for example, where previously PCRE
3905 didn't consider this to be a POSIX class. Likewise for [:1234:].
3906
3907 The problem in trying to be exactly like Perl is in the handling of escapes. We
3908 have to be sure that [abc[:x\]pqr] is *not* treated as containing a POSIX
3909 class, but [abc[:x\]pqr:]] is (so that an error can be generated). The code
3910 below handles the special cases \\ and \], but does not try to do any other
3911 escape processing. This makes it different from Perl for cases such as
3912 [:l\ower:] where Perl recognizes it as the POSIX class "lower" but PCRE does
3913 not recognize "l\ower". This is a lesser evil than not diagnosing bad classes
3914 when Perl does, I think.
3915
3916 A user pointed out that PCRE was rejecting [:a[:digit:]] whereas Perl was not.
3917 It seems that the appearance of a nested POSIX class supersedes an apparent
3918 external class. For example, [:a[:digit:]b:] matches "a", "b", ":", or
3919 a digit.
3920
3921 In Perl, unescaped square brackets may also appear as part of class names. For
3922 example, [:a[:abc]b:] gives unknown POSIX class "[:abc]b:]". However, for
3923 [:a[:abc]b][b:] it gives unknown POSIX class "[:abc]b][b:]", which does not
3924 seem right at all. PCRE does not allow closing square brackets in POSIX class
3925 names.
3926
3927 Arguments:
3928 ptr pointer to the initial [
3929 endptr where to return the end pointer
3930
3931 Returns: TRUE or FALSE
3932 */
3933
3934 static BOOL
check_posix_syntax(const pcre_uchar * ptr,const pcre_uchar ** endptr)3935 check_posix_syntax(const pcre_uchar *ptr, const pcre_uchar **endptr)
3936 {
3937 pcre_uchar terminator; /* Don't combine these lines; the Solaris cc */
3938 terminator = *(++ptr); /* compiler warns about "non-constant" initializer. */
3939 for (++ptr; *ptr != CHAR_NULL; ptr++)
3940 {
3941 if (*ptr == CHAR_BACKSLASH &&
3942 (ptr[1] == CHAR_RIGHT_SQUARE_BRACKET ||
3943 ptr[1] == CHAR_BACKSLASH))
3944 ptr++;
3945 else if ((*ptr == CHAR_LEFT_SQUARE_BRACKET && ptr[1] == terminator) ||
3946 *ptr == CHAR_RIGHT_SQUARE_BRACKET) return FALSE;
3947 else if (*ptr == terminator && ptr[1] == CHAR_RIGHT_SQUARE_BRACKET)
3948 {
3949 *endptr = ptr;
3950 return TRUE;
3951 }
3952 }
3953 return FALSE;
3954 }
3955
3956
3957
3958
3959 /*************************************************
3960 * Check POSIX class name *
3961 *************************************************/
3962
3963 /* This function is called to check the name given in a POSIX-style class entry
3964 such as [:alnum:].
3965
3966 Arguments:
3967 ptr points to the first letter
3968 len the length of the name
3969
3970 Returns: a value representing the name, or -1 if unknown
3971 */
3972
3973 static int
check_posix_name(const pcre_uchar * ptr,int len)3974 check_posix_name(const pcre_uchar *ptr, int len)
3975 {
3976 const char *pn = posix_names;
3977 register int yield = 0;
3978 while (posix_name_lengths[yield] != 0)
3979 {
3980 if (len == posix_name_lengths[yield] &&
3981 STRNCMP_UC_C8(ptr, pn, (unsigned int)len) == 0) return yield;
3982 pn += posix_name_lengths[yield] + 1;
3983 yield++;
3984 }
3985 return -1;
3986 }
3987
3988
3989 /*************************************************
3990 * Adjust OP_RECURSE items in repeated group *
3991 *************************************************/
3992
3993 /* OP_RECURSE items contain an offset from the start of the regex to the group
3994 that is referenced. This means that groups can be replicated for fixed
3995 repetition simply by copying (because the recursion is allowed to refer to
3996 earlier groups that are outside the current group). However, when a group is
3997 optional (i.e. the minimum quantifier is zero), OP_BRAZERO or OP_SKIPZERO is
3998 inserted before it, after it has been compiled. This means that any OP_RECURSE
3999 items within it that refer to the group itself or any contained groups have to
4000 have their offsets adjusted. That one of the jobs of this function. Before it
4001 is called, the partially compiled regex must be temporarily terminated with
4002 OP_END.
4003
4004 This function has been extended to cope with forward references for recursions
4005 and subroutine calls. It must check the list of such references for the
4006 group we are dealing with. If it finds that one of the recursions in the
4007 current group is on this list, it does not adjust the value in the reference
4008 (which is a group number). After the group has been scanned, all the offsets in
4009 the forward reference list for the group are adjusted.
4010
4011 Arguments:
4012 group points to the start of the group
4013 adjust the amount by which the group is to be moved
4014 utf TRUE in UTF-8 / UTF-16 / UTF-32 mode
4015 cd contains pointers to tables etc.
4016 save_hwm_offset the hwm forward reference offset at the start of the group
4017
4018 Returns: nothing
4019 */
4020
4021 static void
adjust_recurse(pcre_uchar * group,int adjust,BOOL utf,compile_data * cd,size_t save_hwm_offset)4022 adjust_recurse(pcre_uchar *group, int adjust, BOOL utf, compile_data *cd,
4023 size_t save_hwm_offset)
4024 {
4025 int offset;
4026 pcre_uchar *hc;
4027 pcre_uchar *ptr = group;
4028
4029 while ((ptr = (pcre_uchar *)find_recurse(ptr, utf)) != NULL)
4030 {
4031 for (hc = (pcre_uchar *)cd->start_workspace + save_hwm_offset; hc < cd->hwm;
4032 hc += LINK_SIZE)
4033 {
4034 offset = (int)GET(hc, 0);
4035 if (cd->start_code + offset == ptr + 1) break;
4036 }
4037
4038 /* If we have not found this recursion on the forward reference list, adjust
4039 the recursion's offset if it's after the start of this group. */
4040
4041 if (hc >= cd->hwm)
4042 {
4043 offset = (int)GET(ptr, 1);
4044 if (cd->start_code + offset >= group) PUT(ptr, 1, offset + adjust);
4045 }
4046
4047 ptr += 1 + LINK_SIZE;
4048 }
4049
4050 /* Now adjust all forward reference offsets for the group. */
4051
4052 for (hc = (pcre_uchar *)cd->start_workspace + save_hwm_offset; hc < cd->hwm;
4053 hc += LINK_SIZE)
4054 {
4055 offset = (int)GET(hc, 0);
4056 PUT(hc, 0, offset + adjust);
4057 }
4058 }
4059
4060
4061
4062 /*************************************************
4063 * Insert an automatic callout point *
4064 *************************************************/
4065
4066 /* This function is called when the PCRE_AUTO_CALLOUT option is set, to insert
4067 callout points before each pattern item.
4068
4069 Arguments:
4070 code current code pointer
4071 ptr current pattern pointer
4072 cd pointers to tables etc
4073
4074 Returns: new code pointer
4075 */
4076
4077 static pcre_uchar *
auto_callout(pcre_uchar * code,const pcre_uchar * ptr,compile_data * cd)4078 auto_callout(pcre_uchar *code, const pcre_uchar *ptr, compile_data *cd)
4079 {
4080 *code++ = OP_CALLOUT;
4081 *code++ = 255;
4082 PUT(code, 0, (int)(ptr - cd->start_pattern)); /* Pattern offset */
4083 PUT(code, LINK_SIZE, 0); /* Default length */
4084 return code + 2 * LINK_SIZE;
4085 }
4086
4087
4088
4089 /*************************************************
4090 * Complete a callout item *
4091 *************************************************/
4092
4093 /* A callout item contains the length of the next item in the pattern, which
4094 we can't fill in till after we have reached the relevant point. This is used
4095 for both automatic and manual callouts.
4096
4097 Arguments:
4098 previous_callout points to previous callout item
4099 ptr current pattern pointer
4100 cd pointers to tables etc
4101
4102 Returns: nothing
4103 */
4104
4105 static void
complete_callout(pcre_uchar * previous_callout,const pcre_uchar * ptr,compile_data * cd)4106 complete_callout(pcre_uchar *previous_callout, const pcre_uchar *ptr, compile_data *cd)
4107 {
4108 int length = (int)(ptr - cd->start_pattern - GET(previous_callout, 2));
4109 PUT(previous_callout, 2 + LINK_SIZE, length);
4110 }
4111
4112
4113
4114 #ifdef SUPPORT_UCP
4115 /*************************************************
4116 * Get othercase range *
4117 *************************************************/
4118
4119 /* This function is passed the start and end of a class range, in UTF-8 mode
4120 with UCP support. It searches up the characters, looking for ranges of
4121 characters in the "other" case. Each call returns the next one, updating the
4122 start address. A character with multiple other cases is returned on its own
4123 with a special return value.
4124
4125 Arguments:
4126 cptr points to starting character value; updated
4127 d end value
4128 ocptr where to put start of othercase range
4129 odptr where to put end of othercase range
4130
4131 Yield: -1 when no more
4132 0 when a range is returned
4133 >0 the CASESET offset for char with multiple other cases
4134 in this case, ocptr contains the original
4135 */
4136
4137 static int
get_othercase_range(pcre_uint32 * cptr,pcre_uint32 d,pcre_uint32 * ocptr,pcre_uint32 * odptr)4138 get_othercase_range(pcre_uint32 *cptr, pcre_uint32 d, pcre_uint32 *ocptr,
4139 pcre_uint32 *odptr)
4140 {
4141 pcre_uint32 c, othercase, next;
4142 unsigned int co;
4143
4144 /* Find the first character that has an other case. If it has multiple other
4145 cases, return its case offset value. */
4146
4147 for (c = *cptr; c <= d; c++)
4148 {
4149 if ((co = UCD_CASESET(c)) != 0)
4150 {
4151 *ocptr = c++; /* Character that has the set */
4152 *cptr = c; /* Rest of input range */
4153 return (int)co;
4154 }
4155 if ((othercase = UCD_OTHERCASE(c)) != c) break;
4156 }
4157
4158 if (c > d) return -1; /* Reached end of range */
4159
4160 /* Found a character that has a single other case. Search for the end of the
4161 range, which is either the end of the input range, or a character that has zero
4162 or more than one other cases. */
4163
4164 *ocptr = othercase;
4165 next = othercase + 1;
4166
4167 for (++c; c <= d; c++)
4168 {
4169 if ((co = UCD_CASESET(c)) != 0 || UCD_OTHERCASE(c) != next) break;
4170 next++;
4171 }
4172
4173 *odptr = next - 1; /* End of othercase range */
4174 *cptr = c; /* Rest of input range */
4175 return 0;
4176 }
4177 #endif /* SUPPORT_UCP */
4178
4179
4180
4181 /*************************************************
4182 * Add a character or range to a class *
4183 *************************************************/
4184
4185 /* This function packages up the logic of adding a character or range of
4186 characters to a class. The character values in the arguments will be within the
4187 valid values for the current mode (8-bit, 16-bit, UTF, etc). This function is
4188 mutually recursive with the function immediately below.
4189
4190 Arguments:
4191 classbits the bit map for characters < 256
4192 uchardptr points to the pointer for extra data
4193 options the options word
4194 cd contains pointers to tables etc.
4195 start start of range character
4196 end end of range character
4197
4198 Returns: the number of < 256 characters added
4199 the pointer to extra data is updated
4200 */
4201
4202 static int
add_to_class(pcre_uint8 * classbits,pcre_uchar ** uchardptr,int options,compile_data * cd,pcre_uint32 start,pcre_uint32 end)4203 add_to_class(pcre_uint8 *classbits, pcre_uchar **uchardptr, int options,
4204 compile_data *cd, pcre_uint32 start, pcre_uint32 end)
4205 {
4206 pcre_uint32 c;
4207 pcre_uint32 classbits_end = (end <= 0xff ? end : 0xff);
4208 int n8 = 0;
4209
4210 /* If caseless matching is required, scan the range and process alternate
4211 cases. In Unicode, there are 8-bit characters that have alternate cases that
4212 are greater than 255 and vice-versa. Sometimes we can just extend the original
4213 range. */
4214
4215 if ((options & PCRE_CASELESS) != 0)
4216 {
4217 #ifdef SUPPORT_UCP
4218 if ((options & PCRE_UTF8) != 0)
4219 {
4220 int rc;
4221 pcre_uint32 oc, od;
4222
4223 options &= ~PCRE_CASELESS; /* Remove for recursive calls */
4224 c = start;
4225
4226 while ((rc = get_othercase_range(&c, end, &oc, &od)) >= 0)
4227 {
4228 /* Handle a single character that has more than one other case. */
4229
4230 if (rc > 0) n8 += add_list_to_class(classbits, uchardptr, options, cd,
4231 PRIV(ucd_caseless_sets) + rc, oc);
4232
4233 /* Do nothing if the other case range is within the original range. */
4234
4235 else if (oc >= start && od <= end) continue;
4236
4237 /* Extend the original range if there is overlap, noting that if oc < c, we
4238 can't have od > end because a subrange is always shorter than the basic
4239 range. Otherwise, use a recursive call to add the additional range. */
4240
4241 else if (oc < start && od >= start - 1) start = oc; /* Extend downwards */
4242 else if (od > end && oc <= end + 1)
4243 {
4244 end = od; /* Extend upwards */
4245 if (end > classbits_end) classbits_end = (end <= 0xff ? end : 0xff);
4246 }
4247 else n8 += add_to_class(classbits, uchardptr, options, cd, oc, od);
4248 }
4249 }
4250 else
4251 #endif /* SUPPORT_UCP */
4252
4253 /* Not UTF-mode, or no UCP */
4254
4255 for (c = start; c <= classbits_end; c++)
4256 {
4257 SETBIT(classbits, cd->fcc[c]);
4258 n8++;
4259 }
4260 }
4261
4262 /* Now handle the original range. Adjust the final value according to the bit
4263 length - this means that the same lists of (e.g.) horizontal spaces can be used
4264 in all cases. */
4265
4266 #if defined COMPILE_PCRE8
4267 #ifdef SUPPORT_UTF
4268 if ((options & PCRE_UTF8) == 0)
4269 #endif
4270 if (end > 0xff) end = 0xff;
4271
4272 #elif defined COMPILE_PCRE16
4273 #ifdef SUPPORT_UTF
4274 if ((options & PCRE_UTF16) == 0)
4275 #endif
4276 if (end > 0xffff) end = 0xffff;
4277
4278 #endif /* COMPILE_PCRE[8|16] */
4279
4280 /* Use the bitmap for characters < 256. Otherwise use extra data.*/
4281
4282 for (c = start; c <= classbits_end; c++)
4283 {
4284 /* Regardless of start, c will always be <= 255. */
4285 SETBIT(classbits, c);
4286 n8++;
4287 }
4288
4289 #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
4290 if (start <= 0xff) start = 0xff + 1;
4291
4292 if (end >= start)
4293 {
4294 pcre_uchar *uchardata = *uchardptr;
4295 #ifdef SUPPORT_UTF
4296 if ((options & PCRE_UTF8) != 0) /* All UTFs use the same flag bit */
4297 {
4298 if (start < end)
4299 {
4300 *uchardata++ = XCL_RANGE;
4301 uchardata += PRIV(ord2utf)(start, uchardata);
4302 uchardata += PRIV(ord2utf)(end, uchardata);
4303 }
4304 else if (start == end)
4305 {
4306 *uchardata++ = XCL_SINGLE;
4307 uchardata += PRIV(ord2utf)(start, uchardata);
4308 }
4309 }
4310 else
4311 #endif /* SUPPORT_UTF */
4312
4313 /* Without UTF support, character values are constrained by the bit length,
4314 and can only be > 256 for 16-bit and 32-bit libraries. */
4315
4316 #ifdef COMPILE_PCRE8
4317 {}
4318 #else
4319 if (start < end)
4320 {
4321 *uchardata++ = XCL_RANGE;
4322 *uchardata++ = start;
4323 *uchardata++ = end;
4324 }
4325 else if (start == end)
4326 {
4327 *uchardata++ = XCL_SINGLE;
4328 *uchardata++ = start;
4329 }
4330 #endif
4331
4332 *uchardptr = uchardata; /* Updata extra data pointer */
4333 }
4334 #endif /* SUPPORT_UTF || !COMPILE_PCRE8 */
4335
4336 return n8; /* Number of 8-bit characters */
4337 }
4338
4339
4340
4341
4342 /*************************************************
4343 * Add a list of characters to a class *
4344 *************************************************/
4345
4346 /* This function is used for adding a list of case-equivalent characters to a
4347 class, and also for adding a list of horizontal or vertical whitespace. If the
4348 list is in order (which it should be), ranges of characters are detected and
4349 handled appropriately. This function is mutually recursive with the function
4350 above.
4351
4352 Arguments:
4353 classbits the bit map for characters < 256
4354 uchardptr points to the pointer for extra data
4355 options the options word
4356 cd contains pointers to tables etc.
4357 p points to row of 32-bit values, terminated by NOTACHAR
4358 except character to omit; this is used when adding lists of
4359 case-equivalent characters to avoid including the one we
4360 already know about
4361
4362 Returns: the number of < 256 characters added
4363 the pointer to extra data is updated
4364 */
4365
4366 static int
add_list_to_class(pcre_uint8 * classbits,pcre_uchar ** uchardptr,int options,compile_data * cd,const pcre_uint32 * p,unsigned int except)4367 add_list_to_class(pcre_uint8 *classbits, pcre_uchar **uchardptr, int options,
4368 compile_data *cd, const pcre_uint32 *p, unsigned int except)
4369 {
4370 int n8 = 0;
4371 while (p[0] < NOTACHAR)
4372 {
4373 int n = 0;
4374 if (p[0] != except)
4375 {
4376 while(p[n+1] == p[0] + n + 1) n++;
4377 n8 += add_to_class(classbits, uchardptr, options, cd, p[0], p[n]);
4378 }
4379 p += n + 1;
4380 }
4381 return n8;
4382 }
4383
4384
4385
4386 /*************************************************
4387 * Add characters not in a list to a class *
4388 *************************************************/
4389
4390 /* This function is used for adding the complement of a list of horizontal or
4391 vertical whitespace to a class. The list must be in order.
4392
4393 Arguments:
4394 classbits the bit map for characters < 256
4395 uchardptr points to the pointer for extra data
4396 options the options word
4397 cd contains pointers to tables etc.
4398 p points to row of 32-bit values, terminated by NOTACHAR
4399
4400 Returns: the number of < 256 characters added
4401 the pointer to extra data is updated
4402 */
4403
4404 static int
add_not_list_to_class(pcre_uint8 * classbits,pcre_uchar ** uchardptr,int options,compile_data * cd,const pcre_uint32 * p)4405 add_not_list_to_class(pcre_uint8 *classbits, pcre_uchar **uchardptr,
4406 int options, compile_data *cd, const pcre_uint32 *p)
4407 {
4408 BOOL utf = (options & PCRE_UTF8) != 0;
4409 int n8 = 0;
4410 if (p[0] > 0)
4411 n8 += add_to_class(classbits, uchardptr, options, cd, 0, p[0] - 1);
4412 while (p[0] < NOTACHAR)
4413 {
4414 while (p[1] == p[0] + 1) p++;
4415 n8 += add_to_class(classbits, uchardptr, options, cd, p[0] + 1,
4416 (p[1] == NOTACHAR) ? (utf ? 0x10ffffu : 0xffffffffu) : p[1] - 1);
4417 p++;
4418 }
4419 return n8;
4420 }
4421
4422
4423
4424 /*************************************************
4425 * Compile one branch *
4426 *************************************************/
4427
4428 /* Scan the pattern, compiling it into the a vector. If the options are
4429 changed during the branch, the pointer is used to change the external options
4430 bits. This function is used during the pre-compile phase when we are trying
4431 to find out the amount of memory needed, as well as during the real compile
4432 phase. The value of lengthptr distinguishes the two phases.
4433
4434 Arguments:
4435 optionsptr pointer to the option bits
4436 codeptr points to the pointer to the current code point
4437 ptrptr points to the current pattern pointer
4438 errorcodeptr points to error code variable
4439 firstcharptr place to put the first required character
4440 firstcharflagsptr place to put the first character flags, or a negative number
4441 reqcharptr place to put the last required character
4442 reqcharflagsptr place to put the last required character flags, or a negative number
4443 bcptr points to current branch chain
4444 cond_depth conditional nesting depth
4445 cd contains pointers to tables etc.
4446 lengthptr NULL during the real compile phase
4447 points to length accumulator during pre-compile phase
4448
4449 Returns: TRUE on success
4450 FALSE, with *errorcodeptr set non-zero on error
4451 */
4452
4453 static BOOL
compile_branch(int * optionsptr,pcre_uchar ** codeptr,const pcre_uchar ** ptrptr,int * errorcodeptr,pcre_uint32 * firstcharptr,pcre_int32 * firstcharflagsptr,pcre_uint32 * reqcharptr,pcre_int32 * reqcharflagsptr,branch_chain * bcptr,int cond_depth,compile_data * cd,int * lengthptr)4454 compile_branch(int *optionsptr, pcre_uchar **codeptr,
4455 const pcre_uchar **ptrptr, int *errorcodeptr,
4456 pcre_uint32 *firstcharptr, pcre_int32 *firstcharflagsptr,
4457 pcre_uint32 *reqcharptr, pcre_int32 *reqcharflagsptr,
4458 branch_chain *bcptr, int cond_depth,
4459 compile_data *cd, int *lengthptr)
4460 {
4461 int repeat_type, op_type;
4462 int repeat_min = 0, repeat_max = 0; /* To please picky compilers */
4463 int bravalue = 0;
4464 int greedy_default, greedy_non_default;
4465 pcre_uint32 firstchar, reqchar;
4466 pcre_int32 firstcharflags, reqcharflags;
4467 pcre_uint32 zeroreqchar, zerofirstchar;
4468 pcre_int32 zeroreqcharflags, zerofirstcharflags;
4469 pcre_int32 req_caseopt, reqvary, tempreqvary;
4470 int options = *optionsptr; /* May change dynamically */
4471 int after_manual_callout = 0;
4472 int length_prevgroup = 0;
4473 register pcre_uint32 c;
4474 int escape;
4475 register pcre_uchar *code = *codeptr;
4476 pcre_uchar *last_code = code;
4477 pcre_uchar *orig_code = code;
4478 pcre_uchar *tempcode;
4479 BOOL inescq = FALSE;
4480 BOOL groupsetfirstchar = FALSE;
4481 const pcre_uchar *ptr = *ptrptr;
4482 const pcre_uchar *tempptr;
4483 const pcre_uchar *nestptr = NULL;
4484 pcre_uchar *previous = NULL;
4485 pcre_uchar *previous_callout = NULL;
4486 size_t item_hwm_offset = 0;
4487 pcre_uint8 classbits[32];
4488
4489 /* We can fish out the UTF-8 setting once and for all into a BOOL, but we
4490 must not do this for other options (e.g. PCRE_EXTENDED) because they may change
4491 dynamically as we process the pattern. */
4492
4493 #ifdef SUPPORT_UTF
4494 /* PCRE_UTF[16|32] have the same value as PCRE_UTF8. */
4495 BOOL utf = (options & PCRE_UTF8) != 0;
4496 #ifndef COMPILE_PCRE32
4497 pcre_uchar utf_chars[6];
4498 #endif
4499 #else
4500 BOOL utf = FALSE;
4501 #endif
4502
4503 /* Helper variables for OP_XCLASS opcode (for characters > 255). We define
4504 class_uchardata always so that it can be passed to add_to_class() always,
4505 though it will not be used in non-UTF 8-bit cases. This avoids having to supply
4506 alternative calls for the different cases. */
4507
4508 pcre_uchar *class_uchardata;
4509 #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
4510 BOOL xclass;
4511 pcre_uchar *class_uchardata_base;
4512 #endif
4513
4514 #ifdef PCRE_DEBUG
4515 if (lengthptr != NULL) DPRINTF((">> start branch\n"));
4516 #endif
4517
4518 /* Set up the default and non-default settings for greediness */
4519
4520 greedy_default = ((options & PCRE_UNGREEDY) != 0);
4521 greedy_non_default = greedy_default ^ 1;
4522
4523 /* Initialize no first byte, no required byte. REQ_UNSET means "no char
4524 matching encountered yet". It gets changed to REQ_NONE if we hit something that
4525 matches a non-fixed char first char; reqchar just remains unset if we never
4526 find one.
4527
4528 When we hit a repeat whose minimum is zero, we may have to adjust these values
4529 to take the zero repeat into account. This is implemented by setting them to
4530 zerofirstbyte and zeroreqchar when such a repeat is encountered. The individual
4531 item types that can be repeated set these backoff variables appropriately. */
4532
4533 firstchar = reqchar = zerofirstchar = zeroreqchar = 0;
4534 firstcharflags = reqcharflags = zerofirstcharflags = zeroreqcharflags = REQ_UNSET;
4535
4536 /* The variable req_caseopt contains either the REQ_CASELESS value
4537 or zero, according to the current setting of the caseless flag. The
4538 REQ_CASELESS leaves the lower 28 bit empty. It is added into the
4539 firstchar or reqchar variables to record the case status of the
4540 value. This is used only for ASCII characters. */
4541
4542 req_caseopt = ((options & PCRE_CASELESS) != 0)? REQ_CASELESS:0;
4543
4544 /* Switch on next character until the end of the branch */
4545
4546 for (;; ptr++)
4547 {
4548 BOOL negate_class;
4549 BOOL should_flip_negation;
4550 BOOL possessive_quantifier;
4551 BOOL is_quantifier;
4552 BOOL is_recurse;
4553 BOOL reset_bracount;
4554 int class_has_8bitchar;
4555 int class_one_char;
4556 #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
4557 BOOL xclass_has_prop;
4558 #endif
4559 int newoptions;
4560 int recno;
4561 int refsign;
4562 int skipbytes;
4563 pcre_uint32 subreqchar, subfirstchar;
4564 pcre_int32 subreqcharflags, subfirstcharflags;
4565 int terminator;
4566 unsigned int mclength;
4567 unsigned int tempbracount;
4568 pcre_uint32 ec;
4569 pcre_uchar mcbuffer[8];
4570
4571 /* Come here to restart the loop without advancing the pointer. */
4572
4573 REDO_LOOP:
4574
4575 /* Get next character in the pattern */
4576
4577 c = *ptr;
4578
4579 /* If we are at the end of a nested substitution, revert to the outer level
4580 string. Nesting only happens one level deep. */
4581
4582 if (c == CHAR_NULL && nestptr != NULL)
4583 {
4584 ptr = nestptr;
4585 nestptr = NULL;
4586 c = *ptr;
4587 }
4588
4589 /* If we are in the pre-compile phase, accumulate the length used for the
4590 previous cycle of this loop. */
4591
4592 if (lengthptr != NULL)
4593 {
4594 #ifdef PCRE_DEBUG
4595 if (code > cd->hwm) cd->hwm = code; /* High water info */
4596 #endif
4597 if (code > cd->start_workspace + cd->workspace_size -
4598 WORK_SIZE_SAFETY_MARGIN) /* Check for overrun */
4599 {
4600 *errorcodeptr = (code >= cd->start_workspace + cd->workspace_size)?
4601 ERR52 : ERR87;
4602 goto FAILED;
4603 }
4604
4605 /* There is at least one situation where code goes backwards: this is the
4606 case of a zero quantifier after a class (e.g. [ab]{0}). At compile time,
4607 the class is simply eliminated. However, it is created first, so we have to
4608 allow memory for it. Therefore, don't ever reduce the length at this point.
4609 */
4610
4611 if (code < last_code) code = last_code;
4612
4613 /* Paranoid check for integer overflow */
4614
4615 if (OFLOW_MAX - *lengthptr < code - last_code)
4616 {
4617 *errorcodeptr = ERR20;
4618 goto FAILED;
4619 }
4620
4621 *lengthptr += (int)(code - last_code);
4622 DPRINTF(("length=%d added %d c=%c (0x%x)\n", *lengthptr,
4623 (int)(code - last_code), c, c));
4624
4625 /* If "previous" is set and it is not at the start of the work space, move
4626 it back to there, in order to avoid filling up the work space. Otherwise,
4627 if "previous" is NULL, reset the current code pointer to the start. */
4628
4629 if (previous != NULL)
4630 {
4631 if (previous > orig_code)
4632 {
4633 memmove(orig_code, previous, IN_UCHARS(code - previous));
4634 code -= previous - orig_code;
4635 previous = orig_code;
4636 }
4637 }
4638 else code = orig_code;
4639
4640 /* Remember where this code item starts so we can pick up the length
4641 next time round. */
4642
4643 last_code = code;
4644 }
4645
4646 /* In the real compile phase, just check the workspace used by the forward
4647 reference list. */
4648
4649 else if (cd->hwm > cd->start_workspace + cd->workspace_size)
4650 {
4651 *errorcodeptr = ERR52;
4652 goto FAILED;
4653 }
4654
4655 /* If in \Q...\E, check for the end; if not, we have a literal. Otherwise an
4656 isolated \E is ignored. */
4657
4658 if (c != CHAR_NULL)
4659 {
4660 if (c == CHAR_BACKSLASH && ptr[1] == CHAR_E)
4661 {
4662 inescq = FALSE;
4663 ptr++;
4664 continue;
4665 }
4666 else if (inescq)
4667 {
4668 if (previous_callout != NULL)
4669 {
4670 if (lengthptr == NULL) /* Don't attempt in pre-compile phase */
4671 complete_callout(previous_callout, ptr, cd);
4672 previous_callout = NULL;
4673 }
4674 if ((options & PCRE_AUTO_CALLOUT) != 0)
4675 {
4676 previous_callout = code;
4677 code = auto_callout(code, ptr, cd);
4678 }
4679 goto NORMAL_CHAR;
4680 }
4681
4682 /* Check for the start of a \Q...\E sequence. We must do this here rather
4683 than later in case it is immediately followed by \E, which turns it into a
4684 "do nothing" sequence. */
4685
4686 if (c == CHAR_BACKSLASH && ptr[1] == CHAR_Q)
4687 {
4688 inescq = TRUE;
4689 ptr++;
4690 continue;
4691 }
4692 }
4693
4694 /* In extended mode, skip white space and comments. */
4695
4696 if ((options & PCRE_EXTENDED) != 0)
4697 {
4698 const pcre_uchar *wscptr = ptr;
4699 while (MAX_255(c) && (cd->ctypes[c] & ctype_space) != 0) c = *(++ptr);
4700 if (c == CHAR_NUMBER_SIGN)
4701 {
4702 ptr++;
4703 while (*ptr != CHAR_NULL)
4704 {
4705 if (IS_NEWLINE(ptr)) /* For non-fixed-length newline cases, */
4706 { /* IS_NEWLINE sets cd->nllen. */
4707 ptr += cd->nllen;
4708 break;
4709 }
4710 ptr++;
4711 #ifdef SUPPORT_UTF
4712 if (utf) FORWARDCHAR(ptr);
4713 #endif
4714 }
4715 }
4716
4717 /* If we skipped any characters, restart the loop. Otherwise, we didn't see
4718 a comment. */
4719
4720 if (ptr > wscptr) goto REDO_LOOP;
4721 }
4722
4723 /* Skip over (?# comments. We need to do this here because we want to know if
4724 the next thing is a quantifier, and these comments may come between an item
4725 and its quantifier. */
4726
4727 if (c == CHAR_LEFT_PARENTHESIS && ptr[1] == CHAR_QUESTION_MARK &&
4728 ptr[2] == CHAR_NUMBER_SIGN)
4729 {
4730 ptr += 3;
4731 while (*ptr != CHAR_NULL && *ptr != CHAR_RIGHT_PARENTHESIS) ptr++;
4732 if (*ptr == CHAR_NULL)
4733 {
4734 *errorcodeptr = ERR18;
4735 goto FAILED;
4736 }
4737 continue;
4738 }
4739
4740 /* See if the next thing is a quantifier. */
4741
4742 is_quantifier =
4743 c == CHAR_ASTERISK || c == CHAR_PLUS || c == CHAR_QUESTION_MARK ||
4744 (c == CHAR_LEFT_CURLY_BRACKET && is_counted_repeat(ptr+1));
4745
4746 /* Fill in length of a previous callout, except when the next thing is a
4747 quantifier or when processing a property substitution string in UCP mode. */
4748
4749 if (!is_quantifier && previous_callout != NULL && nestptr == NULL &&
4750 after_manual_callout-- <= 0)
4751 {
4752 if (lengthptr == NULL) /* Don't attempt in pre-compile phase */
4753 complete_callout(previous_callout, ptr, cd);
4754 previous_callout = NULL;
4755 }
4756
4757 /* Create auto callout, except for quantifiers, or while processing property
4758 strings that are substituted for \w etc in UCP mode. */
4759
4760 if ((options & PCRE_AUTO_CALLOUT) != 0 && !is_quantifier && nestptr == NULL)
4761 {
4762 previous_callout = code;
4763 code = auto_callout(code, ptr, cd);
4764 }
4765
4766 /* Process the next pattern item. */
4767
4768 switch(c)
4769 {
4770 /* ===================================================================*/
4771 case CHAR_NULL: /* The branch terminates at string end */
4772 case CHAR_VERTICAL_LINE: /* or | or ) */
4773 case CHAR_RIGHT_PARENTHESIS:
4774 *firstcharptr = firstchar;
4775 *firstcharflagsptr = firstcharflags;
4776 *reqcharptr = reqchar;
4777 *reqcharflagsptr = reqcharflags;
4778 *codeptr = code;
4779 *ptrptr = ptr;
4780 if (lengthptr != NULL)
4781 {
4782 if (OFLOW_MAX - *lengthptr < code - last_code)
4783 {
4784 *errorcodeptr = ERR20;
4785 goto FAILED;
4786 }
4787 *lengthptr += (int)(code - last_code); /* To include callout length */
4788 DPRINTF((">> end branch\n"));
4789 }
4790 return TRUE;
4791
4792
4793 /* ===================================================================*/
4794 /* Handle single-character metacharacters. In multiline mode, ^ disables
4795 the setting of any following char as a first character. */
4796
4797 case CHAR_CIRCUMFLEX_ACCENT:
4798 previous = NULL;
4799 if ((options & PCRE_MULTILINE) != 0)
4800 {
4801 if (firstcharflags == REQ_UNSET)
4802 zerofirstcharflags = firstcharflags = REQ_NONE;
4803 *code++ = OP_CIRCM;
4804 }
4805 else *code++ = OP_CIRC;
4806 break;
4807
4808 case CHAR_DOLLAR_SIGN:
4809 previous = NULL;
4810 *code++ = ((options & PCRE_MULTILINE) != 0)? OP_DOLLM : OP_DOLL;
4811 break;
4812
4813 /* There can never be a first char if '.' is first, whatever happens about
4814 repeats. The value of reqchar doesn't change either. */
4815
4816 case CHAR_DOT:
4817 if (firstcharflags == REQ_UNSET) firstcharflags = REQ_NONE;
4818 zerofirstchar = firstchar;
4819 zerofirstcharflags = firstcharflags;
4820 zeroreqchar = reqchar;
4821 zeroreqcharflags = reqcharflags;
4822 previous = code;
4823 item_hwm_offset = cd->hwm - cd->start_workspace;
4824 *code++ = ((options & PCRE_DOTALL) != 0)? OP_ALLANY: OP_ANY;
4825 break;
4826
4827
4828 /* ===================================================================*/
4829 /* Character classes. If the included characters are all < 256, we build a
4830 32-byte bitmap of the permitted characters, except in the special case
4831 where there is only one such character. For negated classes, we build the
4832 map as usual, then invert it at the end. However, we use a different opcode
4833 so that data characters > 255 can be handled correctly.
4834
4835 If the class contains characters outside the 0-255 range, a different
4836 opcode is compiled. It may optionally have a bit map for characters < 256,
4837 but those above are are explicitly listed afterwards. A flag byte tells
4838 whether the bitmap is present, and whether this is a negated class or not.
4839
4840 In JavaScript compatibility mode, an isolated ']' causes an error. In
4841 default (Perl) mode, it is treated as a data character. */
4842
4843 case CHAR_RIGHT_SQUARE_BRACKET:
4844 if ((cd->external_options & PCRE_JAVASCRIPT_COMPAT) != 0)
4845 {
4846 *errorcodeptr = ERR64;
4847 goto FAILED;
4848 }
4849 goto NORMAL_CHAR;
4850
4851 /* In another (POSIX) regex library, the ugly syntax [[:<:]] and [[:>:]] is
4852 used for "start of word" and "end of word". As these are otherwise illegal
4853 sequences, we don't break anything by recognizing them. They are replaced
4854 by \b(?=\w) and \b(?<=\w) respectively. Sequences like [a[:<:]] are
4855 erroneous and are handled by the normal code below. */
4856
4857 case CHAR_LEFT_SQUARE_BRACKET:
4858 if (STRNCMP_UC_C8(ptr+1, STRING_WEIRD_STARTWORD, 6) == 0)
4859 {
4860 nestptr = ptr + 7;
4861 ptr = sub_start_of_word;
4862 goto REDO_LOOP;
4863 }
4864
4865 if (STRNCMP_UC_C8(ptr+1, STRING_WEIRD_ENDWORD, 6) == 0)
4866 {
4867 nestptr = ptr + 7;
4868 ptr = sub_end_of_word;
4869 goto REDO_LOOP;
4870 }
4871
4872 /* Handle a real character class. */
4873
4874 previous = code;
4875 item_hwm_offset = cd->hwm - cd->start_workspace;
4876
4877 /* PCRE supports POSIX class stuff inside a class. Perl gives an error if
4878 they are encountered at the top level, so we'll do that too. */
4879
4880 if ((ptr[1] == CHAR_COLON || ptr[1] == CHAR_DOT ||
4881 ptr[1] == CHAR_EQUALS_SIGN) &&
4882 check_posix_syntax(ptr, &tempptr))
4883 {
4884 *errorcodeptr = (ptr[1] == CHAR_COLON)? ERR13 : ERR31;
4885 goto FAILED;
4886 }
4887
4888 /* If the first character is '^', set the negation flag and skip it. Also,
4889 if the first few characters (either before or after ^) are \Q\E or \E we
4890 skip them too. This makes for compatibility with Perl. */
4891
4892 negate_class = FALSE;
4893 for (;;)
4894 {
4895 c = *(++ptr);
4896 if (c == CHAR_BACKSLASH)
4897 {
4898 if (ptr[1] == CHAR_E)
4899 ptr++;
4900 else if (STRNCMP_UC_C8(ptr + 1, STR_Q STR_BACKSLASH STR_E, 3) == 0)
4901 ptr += 3;
4902 else
4903 break;
4904 }
4905 else if (!negate_class && c == CHAR_CIRCUMFLEX_ACCENT)
4906 negate_class = TRUE;
4907 else break;
4908 }
4909
4910 /* Empty classes are allowed in JavaScript compatibility mode. Otherwise,
4911 an initial ']' is taken as a data character -- the code below handles
4912 that. In JS mode, [] must always fail, so generate OP_FAIL, whereas
4913 [^] must match any character, so generate OP_ALLANY. */
4914
4915 if (c == CHAR_RIGHT_SQUARE_BRACKET &&
4916 (cd->external_options & PCRE_JAVASCRIPT_COMPAT) != 0)
4917 {
4918 *code++ = negate_class? OP_ALLANY : OP_FAIL;
4919 if (firstcharflags == REQ_UNSET) firstcharflags = REQ_NONE;
4920 zerofirstchar = firstchar;
4921 zerofirstcharflags = firstcharflags;
4922 break;
4923 }
4924
4925 /* If a class contains a negative special such as \S, we need to flip the
4926 negation flag at the end, so that support for characters > 255 works
4927 correctly (they are all included in the class). */
4928
4929 should_flip_negation = FALSE;
4930
4931 /* Extended class (xclass) will be used when characters > 255
4932 might match. */
4933
4934 #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
4935 xclass = FALSE;
4936 class_uchardata = code + LINK_SIZE + 2; /* For XCLASS items */
4937 class_uchardata_base = class_uchardata; /* Save the start */
4938 #endif
4939
4940 /* For optimization purposes, we track some properties of the class:
4941 class_has_8bitchar will be non-zero if the class contains at least one <
4942 256 character; class_one_char will be 1 if the class contains just one
4943 character; xclass_has_prop will be TRUE if unicode property checks
4944 are present in the class. */
4945
4946 class_has_8bitchar = 0;
4947 class_one_char = 0;
4948 #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
4949 xclass_has_prop = FALSE;
4950 #endif
4951
4952 /* Initialize the 32-char bit map to all zeros. We build the map in a
4953 temporary bit of memory, in case the class contains fewer than two
4954 8-bit characters because in that case the compiled code doesn't use the bit
4955 map. */
4956
4957 memset(classbits, 0, 32 * sizeof(pcre_uint8));
4958
4959 /* Process characters until ] is reached. By writing this as a "do" it
4960 means that an initial ] is taken as a data character. At the start of the
4961 loop, c contains the first byte of the character. */
4962
4963 if (c != CHAR_NULL) do
4964 {
4965 const pcre_uchar *oldptr;
4966
4967 #ifdef SUPPORT_UTF
4968 if (utf && HAS_EXTRALEN(c))
4969 { /* Braces are required because the */
4970 GETCHARLEN(c, ptr, ptr); /* macro generates multiple statements */
4971 }
4972 #endif
4973
4974 #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
4975 /* In the pre-compile phase, accumulate the length of any extra
4976 data and reset the pointer. This is so that very large classes that
4977 contain a zillion > 255 characters no longer overwrite the work space
4978 (which is on the stack). We have to remember that there was XCLASS data,
4979 however. */
4980
4981 if (class_uchardata > class_uchardata_base) xclass = TRUE;
4982
4983 if (lengthptr != NULL && class_uchardata > class_uchardata_base)
4984 {
4985 *lengthptr += (int)(class_uchardata - class_uchardata_base);
4986 class_uchardata = class_uchardata_base;
4987 }
4988 #endif
4989
4990 /* Inside \Q...\E everything is literal except \E */
4991
4992 if (inescq)
4993 {
4994 if (c == CHAR_BACKSLASH && ptr[1] == CHAR_E) /* If we are at \E */
4995 {
4996 inescq = FALSE; /* Reset literal state */
4997 ptr++; /* Skip the 'E' */
4998 continue; /* Carry on with next */
4999 }
5000 goto CHECK_RANGE; /* Could be range if \E follows */
5001 }
5002
5003 /* Handle POSIX class names. Perl allows a negation extension of the
5004 form [:^name:]. A square bracket that doesn't match the syntax is
5005 treated as a literal. We also recognize the POSIX constructions
5006 [.ch.] and [=ch=] ("collating elements") and fault them, as Perl
5007 5.6 and 5.8 do. */
5008
5009 if (c == CHAR_LEFT_SQUARE_BRACKET &&
5010 (ptr[1] == CHAR_COLON || ptr[1] == CHAR_DOT ||
5011 ptr[1] == CHAR_EQUALS_SIGN) && check_posix_syntax(ptr, &tempptr))
5012 {
5013 BOOL local_negate = FALSE;
5014 int posix_class, taboffset, tabopt;
5015 register const pcre_uint8 *cbits = cd->cbits;
5016 pcre_uint8 pbits[32];
5017
5018 if (ptr[1] != CHAR_COLON)
5019 {
5020 *errorcodeptr = ERR31;
5021 goto FAILED;
5022 }
5023
5024 ptr += 2;
5025 if (*ptr == CHAR_CIRCUMFLEX_ACCENT)
5026 {
5027 local_negate = TRUE;
5028 should_flip_negation = TRUE; /* Note negative special */
5029 ptr++;
5030 }
5031
5032 posix_class = check_posix_name(ptr, (int)(tempptr - ptr));
5033 if (posix_class < 0)
5034 {
5035 *errorcodeptr = ERR30;
5036 goto FAILED;
5037 }
5038
5039 /* If matching is caseless, upper and lower are converted to
5040 alpha. This relies on the fact that the class table starts with
5041 alpha, lower, upper as the first 3 entries. */
5042
5043 if ((options & PCRE_CASELESS) != 0 && posix_class <= 2)
5044 posix_class = 0;
5045
5046 /* When PCRE_UCP is set, some of the POSIX classes are converted to
5047 different escape sequences that use Unicode properties \p or \P. Others
5048 that are not available via \p or \P generate XCL_PROP/XCL_NOTPROP
5049 directly. */
5050
5051 #ifdef SUPPORT_UCP
5052 if ((options & PCRE_UCP) != 0)
5053 {
5054 unsigned int ptype = 0;
5055 int pc = posix_class + ((local_negate)? POSIX_SUBSIZE/2 : 0);
5056
5057 /* The posix_substitutes table specifies which POSIX classes can be
5058 converted to \p or \P items. */
5059
5060 if (posix_substitutes[pc] != NULL)
5061 {
5062 nestptr = tempptr + 1;
5063 ptr = posix_substitutes[pc] - 1;
5064 continue;
5065 }
5066
5067 /* There are three other classes that generate special property calls
5068 that are recognized only in an XCLASS. */
5069
5070 else switch(posix_class)
5071 {
5072 case PC_GRAPH:
5073 ptype = PT_PXGRAPH;
5074 /* Fall through */
5075 case PC_PRINT:
5076 if (ptype == 0) ptype = PT_PXPRINT;
5077 /* Fall through */
5078 case PC_PUNCT:
5079 if (ptype == 0) ptype = PT_PXPUNCT;
5080 *class_uchardata++ = local_negate? XCL_NOTPROP : XCL_PROP;
5081 *class_uchardata++ = ptype;
5082 *class_uchardata++ = 0;
5083 xclass_has_prop = TRUE;
5084 ptr = tempptr + 1;
5085 continue;
5086
5087 /* For the other POSIX classes (ascii, cntrl, xdigit) we are going
5088 to fall through to the non-UCP case and build a bit map for
5089 characters with code points less than 256. If we are in a negated
5090 POSIX class, characters with code points greater than 255 must
5091 either all match or all not match. In the special case where we
5092 have not yet generated any xclass data, and this is the final item
5093 in the overall class, we need do nothing: later on, the opcode
5094 OP_NCLASS will be used to indicate that characters greater than 255
5095 are acceptable. If we have already seen an xclass item or one may
5096 follow (we have to assume that it might if this is not the end of
5097 the class), explicitly list all wide codepoints, which will then
5098 either not match or match, depending on whether the class is or is
5099 not negated. */
5100
5101 /* fall through */
5102
5103 default:
5104 if (local_negate &&
5105 (xclass || tempptr[2] != CHAR_RIGHT_SQUARE_BRACKET))
5106 {
5107 *class_uchardata++ = XCL_RANGE;
5108 class_uchardata += PRIV(ord2utf)(0x100, class_uchardata);
5109 class_uchardata += PRIV(ord2utf)(0x10ffff, class_uchardata);
5110 }
5111 break;
5112 }
5113 }
5114 #endif
5115 /* In the non-UCP case, or when UCP makes no difference, we build the
5116 bit map for the POSIX class in a chunk of local store because we may be
5117 adding and subtracting from it, and we don't want to subtract bits that
5118 may be in the main map already. At the end we or the result into the
5119 bit map that is being built. */
5120
5121 posix_class *= 3;
5122
5123 /* Copy in the first table (always present) */
5124
5125 memcpy(pbits, cbits + posix_class_maps[posix_class],
5126 32 * sizeof(pcre_uint8));
5127
5128 /* If there is a second table, add or remove it as required. */
5129
5130 taboffset = posix_class_maps[posix_class + 1];
5131 tabopt = posix_class_maps[posix_class + 2];
5132
5133 if (taboffset >= 0)
5134 {
5135 if (tabopt >= 0)
5136 for (c = 0; c < 32; c++) pbits[c] |= cbits[c + taboffset];
5137 else
5138 for (c = 0; c < 32; c++) pbits[c] &= ~cbits[c + taboffset];
5139 }
5140
5141 /* Now see if we need to remove any special characters. An option
5142 value of 1 removes vertical space and 2 removes underscore. */
5143
5144 if (tabopt < 0) tabopt = -tabopt;
5145 if (tabopt == 1) pbits[1] &= ~0x3c;
5146 else if (tabopt == 2) pbits[11] &= 0x7f;
5147
5148 /* Add the POSIX table or its complement into the main table that is
5149 being built and we are done. */
5150
5151 if (local_negate)
5152 for (c = 0; c < 32; c++) classbits[c] |= ~pbits[c];
5153 else
5154 for (c = 0; c < 32; c++) classbits[c] |= pbits[c];
5155
5156 ptr = tempptr + 1;
5157 /* Every class contains at least one < 256 character. */
5158 class_has_8bitchar = 1;
5159 /* Every class contains at least two characters. */
5160 class_one_char = 2;
5161 continue; /* End of POSIX syntax handling */
5162 }
5163
5164 /* Backslash may introduce a single character, or it may introduce one
5165 of the specials, which just set a flag. The sequence \b is a special
5166 case. Inside a class (and only there) it is treated as backspace. We
5167 assume that other escapes have more than one character in them, so
5168 speculatively set both class_has_8bitchar and class_one_char bigger
5169 than one. Unrecognized escapes fall through and are either treated
5170 as literal characters (by default), or are faulted if
5171 PCRE_EXTRA is set. */
5172
5173 if (c == CHAR_BACKSLASH)
5174 {
5175 escape = check_escape(&ptr, &ec, errorcodeptr, cd->bracount, options,
5176 TRUE);
5177 if (*errorcodeptr != 0) goto FAILED;
5178 if (escape == 0) c = ec;
5179 else if (escape == ESC_b) c = CHAR_BS; /* \b is backspace in a class */
5180 else if (escape == ESC_N) /* \N is not supported in a class */
5181 {
5182 *errorcodeptr = ERR71;
5183 goto FAILED;
5184 }
5185 else if (escape == ESC_Q) /* Handle start of quoted string */
5186 {
5187 if (ptr[1] == CHAR_BACKSLASH && ptr[2] == CHAR_E)
5188 {
5189 ptr += 2; /* avoid empty string */
5190 }
5191 else inescq = TRUE;
5192 continue;
5193 }
5194 else if (escape == ESC_E) continue; /* Ignore orphan \E */
5195
5196 else
5197 {
5198 register const pcre_uint8 *cbits = cd->cbits;
5199 /* Every class contains at least two < 256 characters. */
5200 class_has_8bitchar++;
5201 /* Every class contains at least two characters. */
5202 class_one_char += 2;
5203
5204 switch (escape)
5205 {
5206 #ifdef SUPPORT_UCP
5207 case ESC_du: /* These are the values given for \d etc */
5208 case ESC_DU: /* when PCRE_UCP is set. We replace the */
5209 case ESC_wu: /* escape sequence with an appropriate \p */
5210 case ESC_WU: /* or \P to test Unicode properties instead */
5211 case ESC_su: /* of the default ASCII testing. */
5212 case ESC_SU:
5213 nestptr = ptr;
5214 ptr = substitutes[escape - ESC_DU] - 1; /* Just before substitute */
5215 class_has_8bitchar--; /* Undo! */
5216 continue;
5217 #endif
5218 case ESC_d:
5219 for (c = 0; c < 32; c++) classbits[c] |= cbits[c+cbit_digit];
5220 continue;
5221
5222 case ESC_D:
5223 should_flip_negation = TRUE;
5224 for (c = 0; c < 32; c++) classbits[c] |= ~cbits[c+cbit_digit];
5225 continue;
5226
5227 case ESC_w:
5228 for (c = 0; c < 32; c++) classbits[c] |= cbits[c+cbit_word];
5229 continue;
5230
5231 case ESC_W:
5232 should_flip_negation = TRUE;
5233 for (c = 0; c < 32; c++) classbits[c] |= ~cbits[c+cbit_word];
5234 continue;
5235
5236 /* Perl 5.004 onwards omitted VT from \s, but restored it at Perl
5237 5.18. Before PCRE 8.34, we had to preserve the VT bit if it was
5238 previously set by something earlier in the character class.
5239 Luckily, the value of CHAR_VT is 0x0b in both ASCII and EBCDIC, so
5240 we could just adjust the appropriate bit. From PCRE 8.34 we no
5241 longer treat \s and \S specially. */
5242
5243 case ESC_s:
5244 for (c = 0; c < 32; c++) classbits[c] |= cbits[c+cbit_space];
5245 continue;
5246
5247 case ESC_S:
5248 should_flip_negation = TRUE;
5249 for (c = 0; c < 32; c++) classbits[c] |= ~cbits[c+cbit_space];
5250 continue;
5251
5252 /* The rest apply in both UCP and non-UCP cases. */
5253
5254 case ESC_h:
5255 (void)add_list_to_class(classbits, &class_uchardata, options, cd,
5256 PRIV(hspace_list), NOTACHAR);
5257 continue;
5258
5259 case ESC_H:
5260 (void)add_not_list_to_class(classbits, &class_uchardata, options,
5261 cd, PRIV(hspace_list));
5262 continue;
5263
5264 case ESC_v:
5265 (void)add_list_to_class(classbits, &class_uchardata, options, cd,
5266 PRIV(vspace_list), NOTACHAR);
5267 continue;
5268
5269 case ESC_V:
5270 (void)add_not_list_to_class(classbits, &class_uchardata, options,
5271 cd, PRIV(vspace_list));
5272 continue;
5273
5274 case ESC_p:
5275 case ESC_P:
5276 #ifdef SUPPORT_UCP
5277 {
5278 BOOL negated;
5279 unsigned int ptype = 0, pdata = 0;
5280 if (!get_ucp(&ptr, &negated, &ptype, &pdata, errorcodeptr))
5281 goto FAILED;
5282 *class_uchardata++ = ((escape == ESC_p) != negated)?
5283 XCL_PROP : XCL_NOTPROP;
5284 *class_uchardata++ = ptype;
5285 *class_uchardata++ = pdata;
5286 xclass_has_prop = TRUE;
5287 class_has_8bitchar--; /* Undo! */
5288 continue;
5289 }
5290 #else
5291 *errorcodeptr = ERR45;
5292 goto FAILED;
5293 #endif
5294 /* Unrecognized escapes are faulted if PCRE is running in its
5295 strict mode. By default, for compatibility with Perl, they are
5296 treated as literals. */
5297
5298 default:
5299 if ((options & PCRE_EXTRA) != 0)
5300 {
5301 *errorcodeptr = ERR7;
5302 goto FAILED;
5303 }
5304 class_has_8bitchar--; /* Undo the speculative increase. */
5305 class_one_char -= 2; /* Undo the speculative increase. */
5306 c = *ptr; /* Get the final character and fall through */
5307 break;
5308 }
5309 }
5310
5311 /* Fall through if the escape just defined a single character (c >= 0).
5312 This may be greater than 256. */
5313
5314 escape = 0;
5315
5316 } /* End of backslash handling */
5317
5318 /* A character may be followed by '-' to form a range. However, Perl does
5319 not permit ']' to be the end of the range. A '-' character at the end is
5320 treated as a literal. Perl ignores orphaned \E sequences entirely. The
5321 code for handling \Q and \E is messy. */
5322
5323 CHECK_RANGE:
5324 while (ptr[1] == CHAR_BACKSLASH && ptr[2] == CHAR_E)
5325 {
5326 inescq = FALSE;
5327 ptr += 2;
5328 }
5329 oldptr = ptr;
5330
5331 /* Remember if \r or \n were explicitly used */
5332
5333 if (c == CHAR_CR || c == CHAR_NL) cd->external_flags |= PCRE_HASCRORLF;
5334
5335 /* Check for range */
5336
5337 if (!inescq && ptr[1] == CHAR_MINUS)
5338 {
5339 pcre_uint32 d;
5340 ptr += 2;
5341 while (*ptr == CHAR_BACKSLASH && ptr[1] == CHAR_E) ptr += 2;
5342
5343 /* If we hit \Q (not followed by \E) at this point, go into escaped
5344 mode. */
5345
5346 while (*ptr == CHAR_BACKSLASH && ptr[1] == CHAR_Q)
5347 {
5348 ptr += 2;
5349 if (*ptr == CHAR_BACKSLASH && ptr[1] == CHAR_E)
5350 { ptr += 2; continue; }
5351 inescq = TRUE;
5352 break;
5353 }
5354
5355 /* Minus (hyphen) at the end of a class is treated as a literal, so put
5356 back the pointer and jump to handle the character that preceded it. */
5357
5358 if (*ptr == CHAR_NULL || (!inescq && *ptr == CHAR_RIGHT_SQUARE_BRACKET))
5359 {
5360 ptr = oldptr;
5361 goto CLASS_SINGLE_CHARACTER;
5362 }
5363
5364 /* Otherwise, we have a potential range; pick up the next character */
5365
5366 #ifdef SUPPORT_UTF
5367 if (utf)
5368 { /* Braces are required because the */
5369 GETCHARLEN(d, ptr, ptr); /* macro generates multiple statements */
5370 }
5371 else
5372 #endif
5373 d = *ptr; /* Not UTF-8 mode */
5374
5375 /* The second part of a range can be a single-character escape
5376 sequence, but not any of the other escapes. Perl treats a hyphen as a
5377 literal in such circumstances. However, in Perl's warning mode, a
5378 warning is given, so PCRE now faults it as it is almost certainly a
5379 mistake on the user's part. */
5380
5381 if (!inescq)
5382 {
5383 if (d == CHAR_BACKSLASH)
5384 {
5385 int descape;
5386 descape = check_escape(&ptr, &d, errorcodeptr, cd->bracount, options, TRUE);
5387 if (*errorcodeptr != 0) goto FAILED;
5388
5389 /* 0 means a character was put into d; \b is backspace; any other
5390 special causes an error. */
5391
5392 if (descape != 0)
5393 {
5394 if (descape == ESC_b) d = CHAR_BS; else
5395 {
5396 *errorcodeptr = ERR83;
5397 goto FAILED;
5398 }
5399 }
5400 }
5401
5402 /* A hyphen followed by a POSIX class is treated in the same way. */
5403
5404 else if (d == CHAR_LEFT_SQUARE_BRACKET &&
5405 (ptr[1] == CHAR_COLON || ptr[1] == CHAR_DOT ||
5406 ptr[1] == CHAR_EQUALS_SIGN) &&
5407 check_posix_syntax(ptr, &tempptr))
5408 {
5409 *errorcodeptr = ERR83;
5410 goto FAILED;
5411 }
5412 }
5413
5414 /* Check that the two values are in the correct order. Optimize
5415 one-character ranges. */
5416
5417 if (d < c)
5418 {
5419 *errorcodeptr = ERR8;
5420 goto FAILED;
5421 }
5422 if (d == c) goto CLASS_SINGLE_CHARACTER; /* A few lines below */
5423
5424 /* We have found a character range, so single character optimizations
5425 cannot be done anymore. Any value greater than 1 indicates that there
5426 is more than one character. */
5427
5428 class_one_char = 2;
5429
5430 /* Remember an explicit \r or \n, and add the range to the class. */
5431
5432 if (d == CHAR_CR || d == CHAR_NL) cd->external_flags |= PCRE_HASCRORLF;
5433
5434 class_has_8bitchar +=
5435 add_to_class(classbits, &class_uchardata, options, cd, c, d);
5436
5437 continue; /* Go get the next char in the class */
5438 }
5439
5440 /* Handle a single character - we can get here for a normal non-escape
5441 char, or after \ that introduces a single character or for an apparent
5442 range that isn't. Only the value 1 matters for class_one_char, so don't
5443 increase it if it is already 2 or more ... just in case there's a class
5444 with a zillion characters in it. */
5445
5446 CLASS_SINGLE_CHARACTER:
5447 if (class_one_char < 2) class_one_char++;
5448
5449 /* If xclass_has_prop is false and class_one_char is 1, we have the first
5450 single character in the class, and there have been no prior ranges, or
5451 XCLASS items generated by escapes. If this is the final character in the
5452 class, we can optimize by turning the item into a 1-character OP_CHAR[I]
5453 if it's positive, or OP_NOT[I] if it's negative. In the positive case, it
5454 can cause firstchar to be set. Otherwise, there can be no first char if
5455 this item is first, whatever repeat count may follow. In the case of
5456 reqchar, save the previous value for reinstating. */
5457
5458 if (!inescq &&
5459 #ifdef SUPPORT_UCP
5460 !xclass_has_prop &&
5461 #endif
5462 class_one_char == 1 && ptr[1] == CHAR_RIGHT_SQUARE_BRACKET)
5463 {
5464 ptr++;
5465 zeroreqchar = reqchar;
5466 zeroreqcharflags = reqcharflags;
5467
5468 if (negate_class)
5469 {
5470 #ifdef SUPPORT_UCP
5471 int d;
5472 #endif
5473 if (firstcharflags == REQ_UNSET) firstcharflags = REQ_NONE;
5474 zerofirstchar = firstchar;
5475 zerofirstcharflags = firstcharflags;
5476
5477 /* For caseless UTF-8 mode when UCP support is available, check
5478 whether this character has more than one other case. If so, generate
5479 a special OP_NOTPROP item instead of OP_NOTI. */
5480
5481 #ifdef SUPPORT_UCP
5482 if (utf && (options & PCRE_CASELESS) != 0 &&
5483 (d = UCD_CASESET(c)) != 0)
5484 {
5485 *code++ = OP_NOTPROP;
5486 *code++ = PT_CLIST;
5487 *code++ = d;
5488 }
5489 else
5490 #endif
5491 /* Char has only one other case, or UCP not available */
5492
5493 {
5494 *code++ = ((options & PCRE_CASELESS) != 0)? OP_NOTI: OP_NOT;
5495 #if defined SUPPORT_UTF && !defined COMPILE_PCRE32
5496 if (utf && c > MAX_VALUE_FOR_SINGLE_CHAR)
5497 code += PRIV(ord2utf)(c, code);
5498 else
5499 #endif
5500 *code++ = c;
5501 }
5502
5503 /* We are finished with this character class */
5504
5505 goto END_CLASS;
5506 }
5507
5508 /* For a single, positive character, get the value into mcbuffer, and
5509 then we can handle this with the normal one-character code. */
5510
5511 #if defined SUPPORT_UTF && !defined COMPILE_PCRE32
5512 if (utf && c > MAX_VALUE_FOR_SINGLE_CHAR)
5513 mclength = PRIV(ord2utf)(c, mcbuffer);
5514 else
5515 #endif
5516 {
5517 mcbuffer[0] = c;
5518 mclength = 1;
5519 }
5520 goto ONE_CHAR;
5521 } /* End of 1-char optimization */
5522
5523 /* There is more than one character in the class, or an XCLASS item
5524 has been generated. Add this character to the class. */
5525
5526 class_has_8bitchar +=
5527 add_to_class(classbits, &class_uchardata, options, cd, c, c);
5528 }
5529
5530 /* Loop until ']' reached. This "while" is the end of the "do" far above.
5531 If we are at the end of an internal nested string, revert to the outer
5532 string. */
5533
5534 while (((c = *(++ptr)) != CHAR_NULL ||
5535 (nestptr != NULL &&
5536 (ptr = nestptr, nestptr = NULL, c = *(++ptr)) != CHAR_NULL)) &&
5537 (c != CHAR_RIGHT_SQUARE_BRACKET || inescq));
5538
5539 /* Check for missing terminating ']' */
5540
5541 if (c == CHAR_NULL)
5542 {
5543 *errorcodeptr = ERR6;
5544 goto FAILED;
5545 }
5546
5547 /* We will need an XCLASS if data has been placed in class_uchardata. In
5548 the second phase this is a sufficient test. However, in the pre-compile
5549 phase, class_uchardata gets emptied to prevent workspace overflow, so it
5550 only if the very last character in the class needs XCLASS will it contain
5551 anything at this point. For this reason, xclass gets set TRUE above when
5552 uchar_classdata is emptied, and that's why this code is the way it is here
5553 instead of just doing a test on class_uchardata below. */
5554
5555 #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
5556 if (class_uchardata > class_uchardata_base) xclass = TRUE;
5557 #endif
5558
5559 /* If this is the first thing in the branch, there can be no first char
5560 setting, whatever the repeat count. Any reqchar setting must remain
5561 unchanged after any kind of repeat. */
5562
5563 if (firstcharflags == REQ_UNSET) firstcharflags = REQ_NONE;
5564 zerofirstchar = firstchar;
5565 zerofirstcharflags = firstcharflags;
5566 zeroreqchar = reqchar;
5567 zeroreqcharflags = reqcharflags;
5568
5569 /* If there are characters with values > 255, we have to compile an
5570 extended class, with its own opcode, unless there was a negated special
5571 such as \S in the class, and PCRE_UCP is not set, because in that case all
5572 characters > 255 are in the class, so any that were explicitly given as
5573 well can be ignored. If (when there are explicit characters > 255 that must
5574 be listed) there are no characters < 256, we can omit the bitmap in the
5575 actual compiled code. */
5576
5577 #ifdef SUPPORT_UTF
5578 if (xclass && (xclass_has_prop || !should_flip_negation ||
5579 (options & PCRE_UCP) != 0))
5580 #elif !defined COMPILE_PCRE8
5581 if (xclass && (xclass_has_prop || !should_flip_negation))
5582 #endif
5583 #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
5584 {
5585 /* For non-UCP wide characters, in a non-negative class containing \S or
5586 similar (should_flip_negation is set), all characters greater than 255
5587 must be in the class. */
5588
5589 if (
5590 #if defined COMPILE_PCRE8
5591 utf &&
5592 #endif
5593 should_flip_negation && !negate_class && (options & PCRE_UCP) == 0)
5594 {
5595 *class_uchardata++ = XCL_RANGE;
5596 if (utf) /* Will always be utf in the 8-bit library */
5597 {
5598 class_uchardata += PRIV(ord2utf)(0x100, class_uchardata);
5599 class_uchardata += PRIV(ord2utf)(0x10ffff, class_uchardata);
5600 }
5601 else /* Can only happen for the 16-bit & 32-bit libraries */
5602 {
5603 #if defined COMPILE_PCRE16
5604 *class_uchardata++ = 0x100;
5605 *class_uchardata++ = 0xffffu;
5606 #elif defined COMPILE_PCRE32
5607 *class_uchardata++ = 0x100;
5608 *class_uchardata++ = 0xffffffffu;
5609 #endif
5610 }
5611 }
5612
5613 *class_uchardata++ = XCL_END; /* Marks the end of extra data */
5614 *code++ = OP_XCLASS;
5615 code += LINK_SIZE;
5616 *code = negate_class? XCL_NOT:0;
5617 if (xclass_has_prop) *code |= XCL_HASPROP;
5618
5619 /* If the map is required, move up the extra data to make room for it;
5620 otherwise just move the code pointer to the end of the extra data. */
5621
5622 if (class_has_8bitchar > 0)
5623 {
5624 *code++ |= XCL_MAP;
5625 memmove(code + (32 / sizeof(pcre_uchar)), code,
5626 IN_UCHARS(class_uchardata - code));
5627 if (negate_class && !xclass_has_prop)
5628 for (c = 0; c < 32; c++) classbits[c] = ~classbits[c];
5629 memcpy(code, classbits, 32);
5630 code = class_uchardata + (32 / sizeof(pcre_uchar));
5631 }
5632 else code = class_uchardata;
5633
5634 /* Now fill in the complete length of the item */
5635
5636 PUT(previous, 1, (int)(code - previous));
5637 break; /* End of class handling */
5638 }
5639
5640 /* Even though any XCLASS list is now discarded, we must allow for
5641 its memory. */
5642
5643 if (lengthptr != NULL)
5644 *lengthptr += (int)(class_uchardata - class_uchardata_base);
5645 #endif
5646
5647 /* If there are no characters > 255, or they are all to be included or
5648 excluded, set the opcode to OP_CLASS or OP_NCLASS, depending on whether the
5649 whole class was negated and whether there were negative specials such as \S
5650 (non-UCP) in the class. Then copy the 32-byte map into the code vector,
5651 negating it if necessary. */
5652
5653 *code++ = (negate_class == should_flip_negation) ? OP_CLASS : OP_NCLASS;
5654 if (lengthptr == NULL) /* Save time in the pre-compile phase */
5655 {
5656 if (negate_class)
5657 for (c = 0; c < 32; c++) classbits[c] = ~classbits[c];
5658 memcpy(code, classbits, 32);
5659 }
5660 code += 32 / sizeof(pcre_uchar);
5661
5662 END_CLASS:
5663 break;
5664
5665
5666 /* ===================================================================*/
5667 /* Various kinds of repeat; '{' is not necessarily a quantifier, but this
5668 has been tested above. */
5669
5670 case CHAR_LEFT_CURLY_BRACKET:
5671 if (!is_quantifier) goto NORMAL_CHAR;
5672 ptr = read_repeat_counts(ptr+1, &repeat_min, &repeat_max, errorcodeptr);
5673 if (*errorcodeptr != 0) goto FAILED;
5674 goto REPEAT;
5675
5676 case CHAR_ASTERISK:
5677 repeat_min = 0;
5678 repeat_max = -1;
5679 goto REPEAT;
5680
5681 case CHAR_PLUS:
5682 repeat_min = 1;
5683 repeat_max = -1;
5684 goto REPEAT;
5685
5686 case CHAR_QUESTION_MARK:
5687 repeat_min = 0;
5688 repeat_max = 1;
5689
5690 REPEAT:
5691 if (previous == NULL)
5692 {
5693 *errorcodeptr = ERR9;
5694 goto FAILED;
5695 }
5696
5697 if (repeat_min == 0)
5698 {
5699 firstchar = zerofirstchar; /* Adjust for zero repeat */
5700 firstcharflags = zerofirstcharflags;
5701 reqchar = zeroreqchar; /* Ditto */
5702 reqcharflags = zeroreqcharflags;
5703 }
5704
5705 /* Remember whether this is a variable length repeat */
5706
5707 reqvary = (repeat_min == repeat_max)? 0 : REQ_VARY;
5708
5709 op_type = 0; /* Default single-char op codes */
5710 possessive_quantifier = FALSE; /* Default not possessive quantifier */
5711
5712 /* Save start of previous item, in case we have to move it up in order to
5713 insert something before it. */
5714
5715 tempcode = previous;
5716
5717 /* Before checking for a possessive quantifier, we must skip over
5718 whitespace and comments in extended mode because Perl allows white space at
5719 this point. */
5720
5721 if ((options & PCRE_EXTENDED) != 0)
5722 {
5723 const pcre_uchar *p = ptr + 1;
5724 for (;;)
5725 {
5726 while (MAX_255(*p) && (cd->ctypes[*p] & ctype_space) != 0) p++;
5727 if (*p != CHAR_NUMBER_SIGN) break;
5728 p++;
5729 while (*p != CHAR_NULL)
5730 {
5731 if (IS_NEWLINE(p)) /* For non-fixed-length newline cases, */
5732 { /* IS_NEWLINE sets cd->nllen. */
5733 p += cd->nllen;
5734 break;
5735 }
5736 p++;
5737 #ifdef SUPPORT_UTF
5738 if (utf) FORWARDCHAR(p);
5739 #endif
5740 } /* Loop for comment characters */
5741 } /* Loop for multiple comments */
5742 ptr = p - 1; /* Character before the next significant one. */
5743 }
5744
5745 /* We also need to skip over (?# comments, which are not dependent on
5746 extended mode. */
5747
5748 if (ptr[1] == CHAR_LEFT_PARENTHESIS && ptr[2] == CHAR_QUESTION_MARK &&
5749 ptr[3] == CHAR_NUMBER_SIGN)
5750 {
5751 ptr += 4;
5752 while (*ptr != CHAR_NULL && *ptr != CHAR_RIGHT_PARENTHESIS) ptr++;
5753 if (*ptr == CHAR_NULL)
5754 {
5755 *errorcodeptr = ERR18;
5756 goto FAILED;
5757 }
5758 }
5759
5760 /* If the next character is '+', we have a possessive quantifier. This
5761 implies greediness, whatever the setting of the PCRE_UNGREEDY option.
5762 If the next character is '?' this is a minimizing repeat, by default,
5763 but if PCRE_UNGREEDY is set, it works the other way round. We change the
5764 repeat type to the non-default. */
5765
5766 if (ptr[1] == CHAR_PLUS)
5767 {
5768 repeat_type = 0; /* Force greedy */
5769 possessive_quantifier = TRUE;
5770 ptr++;
5771 }
5772 else if (ptr[1] == CHAR_QUESTION_MARK)
5773 {
5774 repeat_type = greedy_non_default;
5775 ptr++;
5776 }
5777 else repeat_type = greedy_default;
5778
5779 /* If previous was a recursion call, wrap it in atomic brackets so that
5780 previous becomes the atomic group. All recursions were so wrapped in the
5781 past, but it no longer happens for non-repeated recursions. In fact, the
5782 repeated ones could be re-implemented independently so as not to need this,
5783 but for the moment we rely on the code for repeating groups. */
5784
5785 if (*previous == OP_RECURSE)
5786 {
5787 memmove(previous + 1 + LINK_SIZE, previous, IN_UCHARS(1 + LINK_SIZE));
5788 *previous = OP_ONCE;
5789 PUT(previous, 1, 2 + 2*LINK_SIZE);
5790 previous[2 + 2*LINK_SIZE] = OP_KET;
5791 PUT(previous, 3 + 2*LINK_SIZE, 2 + 2*LINK_SIZE);
5792 code += 2 + 2 * LINK_SIZE;
5793 length_prevgroup = 3 + 3*LINK_SIZE;
5794
5795 /* When actually compiling, we need to check whether this was a forward
5796 reference, and if so, adjust the offset. */
5797
5798 if (lengthptr == NULL && cd->hwm >= cd->start_workspace + LINK_SIZE)
5799 {
5800 int offset = GET(cd->hwm, -LINK_SIZE);
5801 if (offset == previous + 1 - cd->start_code)
5802 PUT(cd->hwm, -LINK_SIZE, offset + 1 + LINK_SIZE);
5803 }
5804 }
5805
5806 /* Now handle repetition for the different types of item. */
5807
5808 /* If previous was a character or negated character match, abolish the item
5809 and generate a repeat item instead. If a char item has a minimum of more
5810 than one, ensure that it is set in reqchar - it might not be if a sequence
5811 such as x{3} is the first thing in a branch because the x will have gone
5812 into firstchar instead. */
5813
5814 if (*previous == OP_CHAR || *previous == OP_CHARI
5815 || *previous == OP_NOT || *previous == OP_NOTI)
5816 {
5817 switch (*previous)
5818 {
5819 default: /* Make compiler happy. */
5820 case OP_CHAR: op_type = OP_STAR - OP_STAR; break;
5821 case OP_CHARI: op_type = OP_STARI - OP_STAR; break;
5822 case OP_NOT: op_type = OP_NOTSTAR - OP_STAR; break;
5823 case OP_NOTI: op_type = OP_NOTSTARI - OP_STAR; break;
5824 }
5825
5826 /* Deal with UTF characters that take up more than one character. It's
5827 easier to write this out separately than try to macrify it. Use c to
5828 hold the length of the character in bytes, plus UTF_LENGTH to flag that
5829 it's a length rather than a small character. */
5830
5831 #if defined SUPPORT_UTF && !defined COMPILE_PCRE32
5832 if (utf && NOT_FIRSTCHAR(code[-1]))
5833 {
5834 pcre_uchar *lastchar = code - 1;
5835 BACKCHAR(lastchar);
5836 c = (int)(code - lastchar); /* Length of UTF-8 character */
5837 memcpy(utf_chars, lastchar, IN_UCHARS(c)); /* Save the char */
5838 c |= UTF_LENGTH; /* Flag c as a length */
5839 }
5840 else
5841 #endif /* SUPPORT_UTF */
5842
5843 /* Handle the case of a single charater - either with no UTF support, or
5844 with UTF disabled, or for a single character UTF character. */
5845 {
5846 c = code[-1];
5847 if (*previous <= OP_CHARI && repeat_min > 1)
5848 {
5849 reqchar = c;
5850 reqcharflags = req_caseopt | cd->req_varyopt;
5851 }
5852 }
5853
5854 goto OUTPUT_SINGLE_REPEAT; /* Code shared with single character types */
5855 }
5856
5857 /* If previous was a character type match (\d or similar), abolish it and
5858 create a suitable repeat item. The code is shared with single-character
5859 repeats by setting op_type to add a suitable offset into repeat_type. Note
5860 the the Unicode property types will be present only when SUPPORT_UCP is
5861 defined, but we don't wrap the little bits of code here because it just
5862 makes it horribly messy. */
5863
5864 else if (*previous < OP_EODN)
5865 {
5866 pcre_uchar *oldcode;
5867 int prop_type, prop_value;
5868 op_type = OP_TYPESTAR - OP_STAR; /* Use type opcodes */
5869 c = *previous;
5870
5871 OUTPUT_SINGLE_REPEAT:
5872 if (*previous == OP_PROP || *previous == OP_NOTPROP)
5873 {
5874 prop_type = previous[1];
5875 prop_value = previous[2];
5876 }
5877 else prop_type = prop_value = -1;
5878
5879 oldcode = code;
5880 code = previous; /* Usually overwrite previous item */
5881
5882 /* If the maximum is zero then the minimum must also be zero; Perl allows
5883 this case, so we do too - by simply omitting the item altogether. */
5884
5885 if (repeat_max == 0) goto END_REPEAT;
5886
5887 /* Combine the op_type with the repeat_type */
5888
5889 repeat_type += op_type;
5890
5891 /* A minimum of zero is handled either as the special case * or ?, or as
5892 an UPTO, with the maximum given. */
5893
5894 if (repeat_min == 0)
5895 {
5896 if (repeat_max == -1) *code++ = OP_STAR + repeat_type;
5897 else if (repeat_max == 1) *code++ = OP_QUERY + repeat_type;
5898 else
5899 {
5900 *code++ = OP_UPTO + repeat_type;
5901 PUT2INC(code, 0, repeat_max);
5902 }
5903 }
5904
5905 /* A repeat minimum of 1 is optimized into some special cases. If the
5906 maximum is unlimited, we use OP_PLUS. Otherwise, the original item is
5907 left in place and, if the maximum is greater than 1, we use OP_UPTO with
5908 one less than the maximum. */
5909
5910 else if (repeat_min == 1)
5911 {
5912 if (repeat_max == -1)
5913 *code++ = OP_PLUS + repeat_type;
5914 else
5915 {
5916 code = oldcode; /* leave previous item in place */
5917 if (repeat_max == 1) goto END_REPEAT;
5918 *code++ = OP_UPTO + repeat_type;
5919 PUT2INC(code, 0, repeat_max - 1);
5920 }
5921 }
5922
5923 /* The case {n,n} is just an EXACT, while the general case {n,m} is
5924 handled as an EXACT followed by an UPTO. */
5925
5926 else
5927 {
5928 *code++ = OP_EXACT + op_type; /* NB EXACT doesn't have repeat_type */
5929 PUT2INC(code, 0, repeat_min);
5930
5931 /* If the maximum is unlimited, insert an OP_STAR. Before doing so,
5932 we have to insert the character for the previous code. For a repeated
5933 Unicode property match, there are two extra bytes that define the
5934 required property. In UTF-8 mode, long characters have their length in
5935 c, with the UTF_LENGTH bit as a flag. */
5936
5937 if (repeat_max < 0)
5938 {
5939 #if defined SUPPORT_UTF && !defined COMPILE_PCRE32
5940 if (utf && (c & UTF_LENGTH) != 0)
5941 {
5942 memcpy(code, utf_chars, IN_UCHARS(c & 7));
5943 code += c & 7;
5944 }
5945 else
5946 #endif
5947 {
5948 *code++ = c;
5949 if (prop_type >= 0)
5950 {
5951 *code++ = prop_type;
5952 *code++ = prop_value;
5953 }
5954 }
5955 *code++ = OP_STAR + repeat_type;
5956 }
5957
5958 /* Else insert an UPTO if the max is greater than the min, again
5959 preceded by the character, for the previously inserted code. If the
5960 UPTO is just for 1 instance, we can use QUERY instead. */
5961
5962 else if (repeat_max != repeat_min)
5963 {
5964 #if defined SUPPORT_UTF && !defined COMPILE_PCRE32
5965 if (utf && (c & UTF_LENGTH) != 0)
5966 {
5967 memcpy(code, utf_chars, IN_UCHARS(c & 7));
5968 code += c & 7;
5969 }
5970 else
5971 #endif
5972 *code++ = c;
5973 if (prop_type >= 0)
5974 {
5975 *code++ = prop_type;
5976 *code++ = prop_value;
5977 }
5978 repeat_max -= repeat_min;
5979
5980 if (repeat_max == 1)
5981 {
5982 *code++ = OP_QUERY + repeat_type;
5983 }
5984 else
5985 {
5986 *code++ = OP_UPTO + repeat_type;
5987 PUT2INC(code, 0, repeat_max);
5988 }
5989 }
5990 }
5991
5992 /* The character or character type itself comes last in all cases. */
5993
5994 #if defined SUPPORT_UTF && !defined COMPILE_PCRE32
5995 if (utf && (c & UTF_LENGTH) != 0)
5996 {
5997 memcpy(code, utf_chars, IN_UCHARS(c & 7));
5998 code += c & 7;
5999 }
6000 else
6001 #endif
6002 *code++ = c;
6003
6004 /* For a repeated Unicode property match, there are two extra bytes that
6005 define the required property. */
6006
6007 #ifdef SUPPORT_UCP
6008 if (prop_type >= 0)
6009 {
6010 *code++ = prop_type;
6011 *code++ = prop_value;
6012 }
6013 #endif
6014 }
6015
6016 /* If previous was a character class or a back reference, we put the repeat
6017 stuff after it, but just skip the item if the repeat was {0,0}. */
6018
6019 else if (*previous == OP_CLASS || *previous == OP_NCLASS ||
6020 #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
6021 *previous == OP_XCLASS ||
6022 #endif
6023 *previous == OP_REF || *previous == OP_REFI ||
6024 *previous == OP_DNREF || *previous == OP_DNREFI)
6025 {
6026 if (repeat_max == 0)
6027 {
6028 code = previous;
6029 goto END_REPEAT;
6030 }
6031
6032 if (repeat_min == 0 && repeat_max == -1)
6033 *code++ = OP_CRSTAR + repeat_type;
6034 else if (repeat_min == 1 && repeat_max == -1)
6035 *code++ = OP_CRPLUS + repeat_type;
6036 else if (repeat_min == 0 && repeat_max == 1)
6037 *code++ = OP_CRQUERY + repeat_type;
6038 else
6039 {
6040 *code++ = OP_CRRANGE + repeat_type;
6041 PUT2INC(code, 0, repeat_min);
6042 if (repeat_max == -1) repeat_max = 0; /* 2-byte encoding for max */
6043 PUT2INC(code, 0, repeat_max);
6044 }
6045 }
6046
6047 /* If previous was a bracket group, we may have to replicate it in certain
6048 cases. Note that at this point we can encounter only the "basic" bracket
6049 opcodes such as BRA and CBRA, as this is the place where they get converted
6050 into the more special varieties such as BRAPOS and SBRA. A test for >=
6051 OP_ASSERT and <= OP_COND includes ASSERT, ASSERT_NOT, ASSERTBACK,
6052 ASSERTBACK_NOT, ONCE, ONCE_NC, BRA, BRAPOS, CBRA, CBRAPOS, and COND.
6053 Originally, PCRE did not allow repetition of assertions, but now it does,
6054 for Perl compatibility. */
6055
6056 else if (*previous >= OP_ASSERT && *previous <= OP_COND)
6057 {
6058 register int i;
6059 int len = (int)(code - previous);
6060 size_t base_hwm_offset = item_hwm_offset;
6061 pcre_uchar *bralink = NULL;
6062 pcre_uchar *brazeroptr = NULL;
6063
6064 /* Repeating a DEFINE group is pointless, but Perl allows the syntax, so
6065 we just ignore the repeat. */
6066
6067 if (*previous == OP_COND && previous[LINK_SIZE+1] == OP_DEF)
6068 goto END_REPEAT;
6069
6070 /* There is no sense in actually repeating assertions. The only potential
6071 use of repetition is in cases when the assertion is optional. Therefore,
6072 if the minimum is greater than zero, just ignore the repeat. If the
6073 maximum is not zero or one, set it to 1. */
6074
6075 if (*previous < OP_ONCE) /* Assertion */
6076 {
6077 if (repeat_min > 0) goto END_REPEAT;
6078 if (repeat_max < 0 || repeat_max > 1) repeat_max = 1;
6079 }
6080
6081 /* The case of a zero minimum is special because of the need to stick
6082 OP_BRAZERO in front of it, and because the group appears once in the
6083 data, whereas in other cases it appears the minimum number of times. For
6084 this reason, it is simplest to treat this case separately, as otherwise
6085 the code gets far too messy. There are several special subcases when the
6086 minimum is zero. */
6087
6088 if (repeat_min == 0)
6089 {
6090 /* If the maximum is also zero, we used to just omit the group from the
6091 output altogether, like this:
6092
6093 ** if (repeat_max == 0)
6094 ** {
6095 ** code = previous;
6096 ** goto END_REPEAT;
6097 ** }
6098
6099 However, that fails when a group or a subgroup within it is referenced
6100 as a subroutine from elsewhere in the pattern, so now we stick in
6101 OP_SKIPZERO in front of it so that it is skipped on execution. As we
6102 don't have a list of which groups are referenced, we cannot do this
6103 selectively.
6104
6105 If the maximum is 1 or unlimited, we just have to stick in the BRAZERO
6106 and do no more at this point. However, we do need to adjust any
6107 OP_RECURSE calls inside the group that refer to the group itself or any
6108 internal or forward referenced group, because the offset is from the
6109 start of the whole regex. Temporarily terminate the pattern while doing
6110 this. */
6111
6112 if (repeat_max <= 1) /* Covers 0, 1, and unlimited */
6113 {
6114 *code = OP_END;
6115 adjust_recurse(previous, 1, utf, cd, item_hwm_offset);
6116 memmove(previous + 1, previous, IN_UCHARS(len));
6117 code++;
6118 if (repeat_max == 0)
6119 {
6120 *previous++ = OP_SKIPZERO;
6121 goto END_REPEAT;
6122 }
6123 brazeroptr = previous; /* Save for possessive optimizing */
6124 *previous++ = OP_BRAZERO + repeat_type;
6125 }
6126
6127 /* If the maximum is greater than 1 and limited, we have to replicate
6128 in a nested fashion, sticking OP_BRAZERO before each set of brackets.
6129 The first one has to be handled carefully because it's the original
6130 copy, which has to be moved up. The remainder can be handled by code
6131 that is common with the non-zero minimum case below. We have to
6132 adjust the value or repeat_max, since one less copy is required. Once
6133 again, we may have to adjust any OP_RECURSE calls inside the group. */
6134
6135 else
6136 {
6137 int offset;
6138 *code = OP_END;
6139 adjust_recurse(previous, 2 + LINK_SIZE, utf, cd, item_hwm_offset);
6140 memmove(previous + 2 + LINK_SIZE, previous, IN_UCHARS(len));
6141 code += 2 + LINK_SIZE;
6142 *previous++ = OP_BRAZERO + repeat_type;
6143 *previous++ = OP_BRA;
6144
6145 /* We chain together the bracket offset fields that have to be
6146 filled in later when the ends of the brackets are reached. */
6147
6148 offset = (bralink == NULL)? 0 : (int)(previous - bralink);
6149 bralink = previous;
6150 PUTINC(previous, 0, offset);
6151 }
6152
6153 repeat_max--;
6154 }
6155
6156 /* If the minimum is greater than zero, replicate the group as many
6157 times as necessary, and adjust the maximum to the number of subsequent
6158 copies that we need. If we set a first char from the group, and didn't
6159 set a required char, copy the latter from the former. If there are any
6160 forward reference subroutine calls in the group, there will be entries on
6161 the workspace list; replicate these with an appropriate increment. */
6162
6163 else
6164 {
6165 if (repeat_min > 1)
6166 {
6167 /* In the pre-compile phase, we don't actually do the replication. We
6168 just adjust the length as if we had. Do some paranoid checks for
6169 potential integer overflow. The INT64_OR_DOUBLE type is a 64-bit
6170 integer type when available, otherwise double. */
6171
6172 if (lengthptr != NULL)
6173 {
6174 int delta = (repeat_min - 1)*length_prevgroup;
6175 if ((INT64_OR_DOUBLE)(repeat_min - 1)*
6176 (INT64_OR_DOUBLE)length_prevgroup >
6177 (INT64_OR_DOUBLE)INT_MAX ||
6178 OFLOW_MAX - *lengthptr < delta)
6179 {
6180 *errorcodeptr = ERR20;
6181 goto FAILED;
6182 }
6183 *lengthptr += delta;
6184 }
6185
6186 /* This is compiling for real. If there is a set first byte for
6187 the group, and we have not yet set a "required byte", set it. Make
6188 sure there is enough workspace for copying forward references before
6189 doing the copy. */
6190
6191 else
6192 {
6193 if (groupsetfirstchar && reqcharflags < 0)
6194 {
6195 reqchar = firstchar;
6196 reqcharflags = firstcharflags;
6197 }
6198
6199 for (i = 1; i < repeat_min; i++)
6200 {
6201 pcre_uchar *hc;
6202 size_t this_hwm_offset = cd->hwm - cd->start_workspace;
6203 memcpy(code, previous, IN_UCHARS(len));
6204
6205 while (cd->hwm > cd->start_workspace + cd->workspace_size -
6206 WORK_SIZE_SAFETY_MARGIN -
6207 (this_hwm_offset - base_hwm_offset))
6208 {
6209 *errorcodeptr = expand_workspace(cd);
6210 if (*errorcodeptr != 0) goto FAILED;
6211 }
6212
6213 for (hc = (pcre_uchar *)cd->start_workspace + base_hwm_offset;
6214 hc < (pcre_uchar *)cd->start_workspace + this_hwm_offset;
6215 hc += LINK_SIZE)
6216 {
6217 PUT(cd->hwm, 0, GET(hc, 0) + len);
6218 cd->hwm += LINK_SIZE;
6219 }
6220 base_hwm_offset = this_hwm_offset;
6221 code += len;
6222 }
6223 }
6224 }
6225
6226 if (repeat_max > 0) repeat_max -= repeat_min;
6227 }
6228
6229 /* This code is common to both the zero and non-zero minimum cases. If
6230 the maximum is limited, it replicates the group in a nested fashion,
6231 remembering the bracket starts on a stack. In the case of a zero minimum,
6232 the first one was set up above. In all cases the repeat_max now specifies
6233 the number of additional copies needed. Again, we must remember to
6234 replicate entries on the forward reference list. */
6235
6236 if (repeat_max >= 0)
6237 {
6238 /* In the pre-compile phase, we don't actually do the replication. We
6239 just adjust the length as if we had. For each repetition we must add 1
6240 to the length for BRAZERO and for all but the last repetition we must
6241 add 2 + 2*LINKSIZE to allow for the nesting that occurs. Do some
6242 paranoid checks to avoid integer overflow. The INT64_OR_DOUBLE type is
6243 a 64-bit integer type when available, otherwise double. */
6244
6245 if (lengthptr != NULL && repeat_max > 0)
6246 {
6247 int delta = repeat_max * (length_prevgroup + 1 + 2 + 2*LINK_SIZE) -
6248 2 - 2*LINK_SIZE; /* Last one doesn't nest */
6249 if ((INT64_OR_DOUBLE)repeat_max *
6250 (INT64_OR_DOUBLE)(length_prevgroup + 1 + 2 + 2*LINK_SIZE)
6251 > (INT64_OR_DOUBLE)INT_MAX ||
6252 OFLOW_MAX - *lengthptr < delta)
6253 {
6254 *errorcodeptr = ERR20;
6255 goto FAILED;
6256 }
6257 *lengthptr += delta;
6258 }
6259
6260 /* This is compiling for real */
6261
6262 else for (i = repeat_max - 1; i >= 0; i--)
6263 {
6264 pcre_uchar *hc;
6265 size_t this_hwm_offset = cd->hwm - cd->start_workspace;
6266
6267 *code++ = OP_BRAZERO + repeat_type;
6268
6269 /* All but the final copy start a new nesting, maintaining the
6270 chain of brackets outstanding. */
6271
6272 if (i != 0)
6273 {
6274 int offset;
6275 *code++ = OP_BRA;
6276 offset = (bralink == NULL)? 0 : (int)(code - bralink);
6277 bralink = code;
6278 PUTINC(code, 0, offset);
6279 }
6280
6281 memcpy(code, previous, IN_UCHARS(len));
6282
6283 /* Ensure there is enough workspace for forward references before
6284 copying them. */
6285
6286 while (cd->hwm > cd->start_workspace + cd->workspace_size -
6287 WORK_SIZE_SAFETY_MARGIN -
6288 (this_hwm_offset - base_hwm_offset))
6289 {
6290 *errorcodeptr = expand_workspace(cd);
6291 if (*errorcodeptr != 0) goto FAILED;
6292 }
6293
6294 for (hc = (pcre_uchar *)cd->start_workspace + base_hwm_offset;
6295 hc < (pcre_uchar *)cd->start_workspace + this_hwm_offset;
6296 hc += LINK_SIZE)
6297 {
6298 PUT(cd->hwm, 0, GET(hc, 0) + len + ((i != 0)? 2+LINK_SIZE : 1));
6299 cd->hwm += LINK_SIZE;
6300 }
6301 base_hwm_offset = this_hwm_offset;
6302 code += len;
6303 }
6304
6305 /* Now chain through the pending brackets, and fill in their length
6306 fields (which are holding the chain links pro tem). */
6307
6308 while (bralink != NULL)
6309 {
6310 int oldlinkoffset;
6311 int offset = (int)(code - bralink + 1);
6312 pcre_uchar *bra = code - offset;
6313 oldlinkoffset = GET(bra, 1);
6314 bralink = (oldlinkoffset == 0)? NULL : bralink - oldlinkoffset;
6315 *code++ = OP_KET;
6316 PUTINC(code, 0, offset);
6317 PUT(bra, 1, offset);
6318 }
6319 }
6320
6321 /* If the maximum is unlimited, set a repeater in the final copy. For
6322 ONCE brackets, that's all we need to do. However, possessively repeated
6323 ONCE brackets can be converted into non-capturing brackets, as the
6324 behaviour of (?:xx)++ is the same as (?>xx)++ and this saves having to
6325 deal with possessive ONCEs specially.
6326
6327 Otherwise, when we are doing the actual compile phase, check to see
6328 whether this group is one that could match an empty string. If so,
6329 convert the initial operator to the S form (e.g. OP_BRA -> OP_SBRA) so
6330 that runtime checking can be done. [This check is also applied to ONCE
6331 groups at runtime, but in a different way.]
6332
6333 Then, if the quantifier was possessive and the bracket is not a
6334 conditional, we convert the BRA code to the POS form, and the KET code to
6335 KETRPOS. (It turns out to be convenient at runtime to detect this kind of
6336 subpattern at both the start and at the end.) The use of special opcodes
6337 makes it possible to reduce greatly the stack usage in pcre_exec(). If
6338 the group is preceded by OP_BRAZERO, convert this to OP_BRAPOSZERO.
6339
6340 Then, if the minimum number of matches is 1 or 0, cancel the possessive
6341 flag so that the default action below, of wrapping everything inside
6342 atomic brackets, does not happen. When the minimum is greater than 1,
6343 there will be earlier copies of the group, and so we still have to wrap
6344 the whole thing. */
6345
6346 else
6347 {
6348 pcre_uchar *ketcode = code - 1 - LINK_SIZE;
6349 pcre_uchar *bracode = ketcode - GET(ketcode, 1);
6350
6351 /* Convert possessive ONCE brackets to non-capturing */
6352
6353 if ((*bracode == OP_ONCE || *bracode == OP_ONCE_NC) &&
6354 possessive_quantifier) *bracode = OP_BRA;
6355
6356 /* For non-possessive ONCE brackets, all we need to do is to
6357 set the KET. */
6358
6359 if (*bracode == OP_ONCE || *bracode == OP_ONCE_NC)
6360 *ketcode = OP_KETRMAX + repeat_type;
6361
6362 /* Handle non-ONCE brackets and possessive ONCEs (which have been
6363 converted to non-capturing above). */
6364
6365 else
6366 {
6367 /* In the compile phase, check for empty string matching. */
6368
6369 if (lengthptr == NULL)
6370 {
6371 pcre_uchar *scode = bracode;
6372 do
6373 {
6374 if (could_be_empty_branch(scode, ketcode, utf, cd, NULL))
6375 {
6376 *bracode += OP_SBRA - OP_BRA;
6377 break;
6378 }
6379 scode += GET(scode, 1);
6380 }
6381 while (*scode == OP_ALT);
6382 }
6383
6384 /* A conditional group with only one branch has an implicit empty
6385 alternative branch. */
6386
6387 if (*bracode == OP_COND && bracode[GET(bracode,1)] != OP_ALT)
6388 *bracode = OP_SCOND;
6389
6390 /* Handle possessive quantifiers. */
6391
6392 if (possessive_quantifier)
6393 {
6394 /* For COND brackets, we wrap the whole thing in a possessively
6395 repeated non-capturing bracket, because we have not invented POS
6396 versions of the COND opcodes. Because we are moving code along, we
6397 must ensure that any pending recursive references are updated. */
6398
6399 if (*bracode == OP_COND || *bracode == OP_SCOND)
6400 {
6401 int nlen = (int)(code - bracode);
6402 *code = OP_END;
6403 adjust_recurse(bracode, 1 + LINK_SIZE, utf, cd, item_hwm_offset);
6404 memmove(bracode + 1 + LINK_SIZE, bracode, IN_UCHARS(nlen));
6405 code += 1 + LINK_SIZE;
6406 nlen += 1 + LINK_SIZE;
6407 *bracode = (*bracode == OP_COND)? OP_BRAPOS : OP_SBRAPOS;
6408 *code++ = OP_KETRPOS;
6409 PUTINC(code, 0, nlen);
6410 PUT(bracode, 1, nlen);
6411 }
6412
6413 /* For non-COND brackets, we modify the BRA code and use KETRPOS. */
6414
6415 else
6416 {
6417 *bracode += 1; /* Switch to xxxPOS opcodes */
6418 *ketcode = OP_KETRPOS;
6419 }
6420
6421 /* If the minimum is zero, mark it as possessive, then unset the
6422 possessive flag when the minimum is 0 or 1. */
6423
6424 if (brazeroptr != NULL) *brazeroptr = OP_BRAPOSZERO;
6425 if (repeat_min < 2) possessive_quantifier = FALSE;
6426 }
6427
6428 /* Non-possessive quantifier */
6429
6430 else *ketcode = OP_KETRMAX + repeat_type;
6431 }
6432 }
6433 }
6434
6435 /* If previous is OP_FAIL, it was generated by an empty class [] in
6436 JavaScript mode. The other ways in which OP_FAIL can be generated, that is
6437 by (*FAIL) or (?!) set previous to NULL, which gives a "nothing to repeat"
6438 error above. We can just ignore the repeat in JS case. */
6439
6440 else if (*previous == OP_FAIL) goto END_REPEAT;
6441
6442 /* Else there's some kind of shambles */
6443
6444 else
6445 {
6446 *errorcodeptr = ERR11;
6447 goto FAILED;
6448 }
6449
6450 /* If the character following a repeat is '+', possessive_quantifier is
6451 TRUE. For some opcodes, there are special alternative opcodes for this
6452 case. For anything else, we wrap the entire repeated item inside OP_ONCE
6453 brackets. Logically, the '+' notation is just syntactic sugar, taken from
6454 Sun's Java package, but the special opcodes can optimize it.
6455
6456 Some (but not all) possessively repeated subpatterns have already been
6457 completely handled in the code just above. For them, possessive_quantifier
6458 is always FALSE at this stage. Note that the repeated item starts at
6459 tempcode, not at previous, which might be the first part of a string whose
6460 (former) last char we repeated. */
6461
6462 if (possessive_quantifier)
6463 {
6464 int len;
6465
6466 /* Possessifying an EXACT quantifier has no effect, so we can ignore it.
6467 However, QUERY, STAR, or UPTO may follow (for quantifiers such as {5,6},
6468 {5,}, or {5,10}). We skip over an EXACT item; if the length of what
6469 remains is greater than zero, there's a further opcode that can be
6470 handled. If not, do nothing, leaving the EXACT alone. */
6471
6472 switch(*tempcode)
6473 {
6474 case OP_TYPEEXACT:
6475 tempcode += PRIV(OP_lengths)[*tempcode] +
6476 ((tempcode[1 + IMM2_SIZE] == OP_PROP
6477 || tempcode[1 + IMM2_SIZE] == OP_NOTPROP)? 2 : 0);
6478 break;
6479
6480 /* CHAR opcodes are used for exacts whose count is 1. */
6481
6482 case OP_CHAR:
6483 case OP_CHARI:
6484 case OP_NOT:
6485 case OP_NOTI:
6486 case OP_EXACT:
6487 case OP_EXACTI:
6488 case OP_NOTEXACT:
6489 case OP_NOTEXACTI:
6490 tempcode += PRIV(OP_lengths)[*tempcode];
6491 #ifdef SUPPORT_UTF
6492 if (utf && HAS_EXTRALEN(tempcode[-1]))
6493 tempcode += GET_EXTRALEN(tempcode[-1]);
6494 #endif
6495 break;
6496
6497 /* For the class opcodes, the repeat operator appears at the end;
6498 adjust tempcode to point to it. */
6499
6500 case OP_CLASS:
6501 case OP_NCLASS:
6502 tempcode += 1 + 32/sizeof(pcre_uchar);
6503 break;
6504
6505 #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
6506 case OP_XCLASS:
6507 tempcode += GET(tempcode, 1);
6508 break;
6509 #endif
6510 }
6511
6512 /* If tempcode is equal to code (which points to the end of the repeated
6513 item), it means we have skipped an EXACT item but there is no following
6514 QUERY, STAR, or UPTO; the value of len will be 0, and we do nothing. In
6515 all other cases, tempcode will be pointing to the repeat opcode, and will
6516 be less than code, so the value of len will be greater than 0. */
6517
6518 len = (int)(code - tempcode);
6519 if (len > 0)
6520 {
6521 unsigned int repcode = *tempcode;
6522
6523 /* There is a table for possessifying opcodes, all of which are less
6524 than OP_CALLOUT. A zero entry means there is no possessified version.
6525 */
6526
6527 if (repcode < OP_CALLOUT && opcode_possessify[repcode] > 0)
6528 *tempcode = opcode_possessify[repcode];
6529
6530 /* For opcode without a special possessified version, wrap the item in
6531 ONCE brackets. Because we are moving code along, we must ensure that any
6532 pending recursive references are updated. */
6533
6534 else
6535 {
6536 *code = OP_END;
6537 adjust_recurse(tempcode, 1 + LINK_SIZE, utf, cd, item_hwm_offset);
6538 memmove(tempcode + 1 + LINK_SIZE, tempcode, IN_UCHARS(len));
6539 code += 1 + LINK_SIZE;
6540 len += 1 + LINK_SIZE;
6541 tempcode[0] = OP_ONCE;
6542 *code++ = OP_KET;
6543 PUTINC(code, 0, len);
6544 PUT(tempcode, 1, len);
6545 }
6546 }
6547
6548 #ifdef NEVER
6549 if (len > 0) switch (*tempcode)
6550 {
6551 case OP_STAR: *tempcode = OP_POSSTAR; break;
6552 case OP_PLUS: *tempcode = OP_POSPLUS; break;
6553 case OP_QUERY: *tempcode = OP_POSQUERY; break;
6554 case OP_UPTO: *tempcode = OP_POSUPTO; break;
6555
6556 case OP_STARI: *tempcode = OP_POSSTARI; break;
6557 case OP_PLUSI: *tempcode = OP_POSPLUSI; break;
6558 case OP_QUERYI: *tempcode = OP_POSQUERYI; break;
6559 case OP_UPTOI: *tempcode = OP_POSUPTOI; break;
6560
6561 case OP_NOTSTAR: *tempcode = OP_NOTPOSSTAR; break;
6562 case OP_NOTPLUS: *tempcode = OP_NOTPOSPLUS; break;
6563 case OP_NOTQUERY: *tempcode = OP_NOTPOSQUERY; break;
6564 case OP_NOTUPTO: *tempcode = OP_NOTPOSUPTO; break;
6565
6566 case OP_NOTSTARI: *tempcode = OP_NOTPOSSTARI; break;
6567 case OP_NOTPLUSI: *tempcode = OP_NOTPOSPLUSI; break;
6568 case OP_NOTQUERYI: *tempcode = OP_NOTPOSQUERYI; break;
6569 case OP_NOTUPTOI: *tempcode = OP_NOTPOSUPTOI; break;
6570
6571 case OP_TYPESTAR: *tempcode = OP_TYPEPOSSTAR; break;
6572 case OP_TYPEPLUS: *tempcode = OP_TYPEPOSPLUS; break;
6573 case OP_TYPEQUERY: *tempcode = OP_TYPEPOSQUERY; break;
6574 case OP_TYPEUPTO: *tempcode = OP_TYPEPOSUPTO; break;
6575
6576 case OP_CRSTAR: *tempcode = OP_CRPOSSTAR; break;
6577 case OP_CRPLUS: *tempcode = OP_CRPOSPLUS; break;
6578 case OP_CRQUERY: *tempcode = OP_CRPOSQUERY; break;
6579 case OP_CRRANGE: *tempcode = OP_CRPOSRANGE; break;
6580
6581 /* Because we are moving code along, we must ensure that any
6582 pending recursive references are updated. */
6583
6584 default:
6585 *code = OP_END;
6586 adjust_recurse(tempcode, 1 + LINK_SIZE, utf, cd, item_hwm_offset);
6587 memmove(tempcode + 1 + LINK_SIZE, tempcode, IN_UCHARS(len));
6588 code += 1 + LINK_SIZE;
6589 len += 1 + LINK_SIZE;
6590 tempcode[0] = OP_ONCE;
6591 *code++ = OP_KET;
6592 PUTINC(code, 0, len);
6593 PUT(tempcode, 1, len);
6594 break;
6595 }
6596 #endif
6597 }
6598
6599 /* In all case we no longer have a previous item. We also set the
6600 "follows varying string" flag for subsequently encountered reqchars if
6601 it isn't already set and we have just passed a varying length item. */
6602
6603 END_REPEAT:
6604 previous = NULL;
6605 cd->req_varyopt |= reqvary;
6606 break;
6607
6608
6609 /* ===================================================================*/
6610 /* Start of nested parenthesized sub-expression, or comment or lookahead or
6611 lookbehind or option setting or condition or all the other extended
6612 parenthesis forms. */
6613
6614 case CHAR_LEFT_PARENTHESIS:
6615 ptr++;
6616
6617 /* Now deal with various "verbs" that can be introduced by '*'. */
6618
6619 if (ptr[0] == CHAR_ASTERISK && (ptr[1] == ':'
6620 || (MAX_255(ptr[1]) && ((cd->ctypes[ptr[1]] & ctype_letter) != 0))))
6621 {
6622 int i, namelen;
6623 int arglen = 0;
6624 const char *vn = verbnames;
6625 const pcre_uchar *name = ptr + 1;
6626 const pcre_uchar *arg = NULL;
6627 previous = NULL;
6628 ptr++;
6629 while (MAX_255(*ptr) && (cd->ctypes[*ptr] & ctype_letter) != 0) ptr++;
6630 namelen = (int)(ptr - name);
6631
6632 /* It appears that Perl allows any characters whatsoever, other than
6633 a closing parenthesis, to appear in arguments, so we no longer insist on
6634 letters, digits, and underscores. */
6635
6636 if (*ptr == CHAR_COLON)
6637 {
6638 arg = ++ptr;
6639 while (*ptr != CHAR_NULL && *ptr != CHAR_RIGHT_PARENTHESIS) ptr++;
6640 arglen = (int)(ptr - arg);
6641 if ((unsigned int)arglen > MAX_MARK)
6642 {
6643 *errorcodeptr = ERR75;
6644 goto FAILED;
6645 }
6646 }
6647
6648 if (*ptr != CHAR_RIGHT_PARENTHESIS)
6649 {
6650 *errorcodeptr = ERR60;
6651 goto FAILED;
6652 }
6653
6654 /* Scan the table of verb names */
6655
6656 for (i = 0; i < verbcount; i++)
6657 {
6658 if (namelen == verbs[i].len &&
6659 STRNCMP_UC_C8(name, vn, namelen) == 0)
6660 {
6661 int setverb;
6662
6663 /* Check for open captures before ACCEPT and convert it to
6664 ASSERT_ACCEPT if in an assertion. */
6665
6666 if (verbs[i].op == OP_ACCEPT)
6667 {
6668 open_capitem *oc;
6669 if (arglen != 0)
6670 {
6671 *errorcodeptr = ERR59;
6672 goto FAILED;
6673 }
6674 cd->had_accept = TRUE;
6675 for (oc = cd->open_caps; oc != NULL; oc = oc->next)
6676 {
6677 if (lengthptr != NULL)
6678 {
6679 #ifdef COMPILE_PCRE8
6680 *lengthptr += 1 + IMM2_SIZE;
6681 #elif defined COMPILE_PCRE16
6682 *lengthptr += 2 + IMM2_SIZE;
6683 #elif defined COMPILE_PCRE32
6684 *lengthptr += 4 + IMM2_SIZE;
6685 #endif
6686 }
6687 else
6688 {
6689 *code++ = OP_CLOSE;
6690 PUT2INC(code, 0, oc->number);
6691 }
6692 }
6693 setverb = *code++ =
6694 (cd->assert_depth > 0)? OP_ASSERT_ACCEPT : OP_ACCEPT;
6695
6696 /* Do not set firstchar after *ACCEPT */
6697 if (firstcharflags == REQ_UNSET) firstcharflags = REQ_NONE;
6698 }
6699
6700 /* Handle other cases with/without an argument */
6701
6702 else if (arglen == 0)
6703 {
6704 if (verbs[i].op < 0) /* Argument is mandatory */
6705 {
6706 *errorcodeptr = ERR66;
6707 goto FAILED;
6708 }
6709 setverb = *code++ = verbs[i].op;
6710 }
6711
6712 else
6713 {
6714 if (verbs[i].op_arg < 0) /* Argument is forbidden */
6715 {
6716 *errorcodeptr = ERR59;
6717 goto FAILED;
6718 }
6719 setverb = *code++ = verbs[i].op_arg;
6720 if (lengthptr != NULL) /* In pass 1 just add in the length */
6721 { /* to avoid potential workspace */
6722 *lengthptr += arglen; /* overflow. */
6723 *code++ = 0;
6724 }
6725 else
6726 {
6727 *code++ = arglen;
6728 memcpy(code, arg, IN_UCHARS(arglen));
6729 code += arglen;
6730 }
6731 *code++ = 0;
6732 }
6733
6734 switch (setverb)
6735 {
6736 case OP_THEN:
6737 case OP_THEN_ARG:
6738 cd->external_flags |= PCRE_HASTHEN;
6739 break;
6740
6741 case OP_PRUNE:
6742 case OP_PRUNE_ARG:
6743 case OP_SKIP:
6744 case OP_SKIP_ARG:
6745 cd->had_pruneorskip = TRUE;
6746 break;
6747 }
6748
6749 break; /* Found verb, exit loop */
6750 }
6751
6752 vn += verbs[i].len + 1;
6753 }
6754
6755 if (i < verbcount) continue; /* Successfully handled a verb */
6756 *errorcodeptr = ERR60; /* Verb not recognized */
6757 goto FAILED;
6758 }
6759
6760 /* Initialize for "real" parentheses */
6761
6762 newoptions = options;
6763 skipbytes = 0;
6764 bravalue = OP_CBRA;
6765 item_hwm_offset = cd->hwm - cd->start_workspace;
6766 reset_bracount = FALSE;
6767
6768 /* Deal with the extended parentheses; all are introduced by '?', and the
6769 appearance of any of them means that this is not a capturing group. */
6770
6771 if (*ptr == CHAR_QUESTION_MARK)
6772 {
6773 int i, set, unset, namelen;
6774 int *optset;
6775 const pcre_uchar *name;
6776 pcre_uchar *slot;
6777
6778 switch (*(++ptr))
6779 {
6780 /* ------------------------------------------------------------ */
6781 case CHAR_VERTICAL_LINE: /* Reset capture count for each branch */
6782 reset_bracount = TRUE;
6783 cd->dupgroups = TRUE; /* Record (?| encountered */
6784 /* Fall through */
6785
6786 /* ------------------------------------------------------------ */
6787 case CHAR_COLON: /* Non-capturing bracket */
6788 bravalue = OP_BRA;
6789 ptr++;
6790 break;
6791
6792
6793 /* ------------------------------------------------------------ */
6794 case CHAR_LEFT_PARENTHESIS:
6795 bravalue = OP_COND; /* Conditional group */
6796 tempptr = ptr;
6797
6798 /* A condition can be an assertion, a number (referring to a numbered
6799 group's having been set), a name (referring to a named group), or 'R',
6800 referring to recursion. R<digits> and R&name are also permitted for
6801 recursion tests.
6802
6803 There are ways of testing a named group: (?(name)) is used by Python;
6804 Perl 5.10 onwards uses (?(<name>) or (?('name')).
6805
6806 There is one unfortunate ambiguity, caused by history. 'R' can be the
6807 recursive thing or the name 'R' (and similarly for 'R' followed by
6808 digits). We look for a name first; if not found, we try the other case.
6809
6810 For compatibility with auto-callouts, we allow a callout to be
6811 specified before a condition that is an assertion. First, check for the
6812 syntax of a callout; if found, adjust the temporary pointer that is
6813 used to check for an assertion condition. That's all that is needed! */
6814
6815 if (ptr[1] == CHAR_QUESTION_MARK && ptr[2] == CHAR_C)
6816 {
6817 for (i = 3;; i++) if (!IS_DIGIT(ptr[i])) break;
6818 if (ptr[i] == CHAR_RIGHT_PARENTHESIS)
6819 tempptr += i + 1;
6820
6821 /* tempptr should now be pointing to the opening parenthesis of the
6822 assertion condition. */
6823
6824 if (*tempptr != CHAR_LEFT_PARENTHESIS)
6825 {
6826 *errorcodeptr = ERR28;
6827 goto FAILED;
6828 }
6829 }
6830
6831 /* For conditions that are assertions, check the syntax, and then exit
6832 the switch. This will take control down to where bracketed groups,
6833 including assertions, are processed. */
6834
6835 if (tempptr[1] == CHAR_QUESTION_MARK &&
6836 (tempptr[2] == CHAR_EQUALS_SIGN ||
6837 tempptr[2] == CHAR_EXCLAMATION_MARK ||
6838 (tempptr[2] == CHAR_LESS_THAN_SIGN &&
6839 (tempptr[3] == CHAR_EQUALS_SIGN ||
6840 tempptr[3] == CHAR_EXCLAMATION_MARK))))
6841 {
6842 cd->iscondassert = TRUE;
6843 break;
6844 }
6845
6846 /* Other conditions use OP_CREF/OP_DNCREF/OP_RREF/OP_DNRREF, and all
6847 need to skip at least 1+IMM2_SIZE bytes at the start of the group. */
6848
6849 code[1+LINK_SIZE] = OP_CREF;
6850 skipbytes = 1+IMM2_SIZE;
6851 refsign = -1; /* => not a number */
6852 namelen = -1; /* => not a name; must set to avoid warning */
6853 name = NULL; /* Always set to avoid warning */
6854 recno = 0; /* Always set to avoid warning */
6855
6856 /* Check for a test for recursion in a named group. */
6857
6858 ptr++;
6859 if (*ptr == CHAR_R && ptr[1] == CHAR_AMPERSAND)
6860 {
6861 terminator = -1;
6862 ptr += 2;
6863 code[1+LINK_SIZE] = OP_RREF; /* Change the type of test */
6864 }
6865
6866 /* Check for a test for a named group's having been set, using the Perl
6867 syntax (?(<name>) or (?('name'), and also allow for the original PCRE
6868 syntax of (?(name) or for (?(+n), (?(-n), and just (?(n). */
6869
6870 else if (*ptr == CHAR_LESS_THAN_SIGN)
6871 {
6872 terminator = CHAR_GREATER_THAN_SIGN;
6873 ptr++;
6874 }
6875 else if (*ptr == CHAR_APOSTROPHE)
6876 {
6877 terminator = CHAR_APOSTROPHE;
6878 ptr++;
6879 }
6880 else
6881 {
6882 terminator = CHAR_NULL;
6883 if (*ptr == CHAR_MINUS || *ptr == CHAR_PLUS) refsign = *ptr++;
6884 else if (IS_DIGIT(*ptr)) refsign = 0;
6885 }
6886
6887 /* Handle a number */
6888
6889 if (refsign >= 0)
6890 {
6891 while (IS_DIGIT(*ptr))
6892 {
6893 if (recno > INT_MAX / 10 - 1) /* Integer overflow */
6894 {
6895 while (IS_DIGIT(*ptr)) ptr++;
6896 *errorcodeptr = ERR61;
6897 goto FAILED;
6898 }
6899 recno = recno * 10 + (int)(*ptr - CHAR_0);
6900 ptr++;
6901 }
6902 }
6903
6904 /* Otherwise we expect to read a name; anything else is an error. When
6905 a name is one of a number of duplicates, a different opcode is used and
6906 it needs more memory. Unfortunately we cannot tell whether a name is a
6907 duplicate in the first pass, so we have to allow for more memory. */
6908
6909 else
6910 {
6911 if (IS_DIGIT(*ptr))
6912 {
6913 *errorcodeptr = ERR84;
6914 goto FAILED;
6915 }
6916 if (!MAX_255(*ptr) || (cd->ctypes[*ptr] & ctype_word) == 0)
6917 {
6918 *errorcodeptr = ERR28; /* Assertion expected */
6919 goto FAILED;
6920 }
6921 name = ptr++;
6922 while (MAX_255(*ptr) && (cd->ctypes[*ptr] & ctype_word) != 0)
6923 {
6924 ptr++;
6925 }
6926 namelen = (int)(ptr - name);
6927 if (lengthptr != NULL) skipbytes += IMM2_SIZE;
6928 }
6929
6930 /* Check the terminator */
6931
6932 if ((terminator > 0 && *ptr++ != (pcre_uchar)terminator) ||
6933 *ptr++ != CHAR_RIGHT_PARENTHESIS)
6934 {
6935 ptr--; /* Error offset */
6936 *errorcodeptr = ERR26; /* Malformed number or name */
6937 goto FAILED;
6938 }
6939
6940 /* Do no further checking in the pre-compile phase. */
6941
6942 if (lengthptr != NULL) break;
6943
6944 /* In the real compile we do the work of looking for the actual
6945 reference. If refsign is not negative, it means we have a number in
6946 recno. */
6947
6948 if (refsign >= 0)
6949 {
6950 if (recno <= 0)
6951 {
6952 *errorcodeptr = ERR35;
6953 goto FAILED;
6954 }
6955 if (refsign != 0) recno = (refsign == CHAR_MINUS)?
6956 cd->bracount - recno + 1 : recno + cd->bracount;
6957 if (recno <= 0 || recno > cd->final_bracount)
6958 {
6959 *errorcodeptr = ERR15;
6960 goto FAILED;
6961 }
6962 PUT2(code, 2+LINK_SIZE, recno);
6963 if (recno > cd->top_backref) cd->top_backref = recno;
6964 break;
6965 }
6966
6967 /* Otherwise look for the name. */
6968
6969 slot = cd->name_table;
6970 for (i = 0; i < cd->names_found; i++)
6971 {
6972 if (STRNCMP_UC_UC(name, slot+IMM2_SIZE, namelen) == 0 &&
6973 slot[IMM2_SIZE+namelen] == 0) break;
6974 slot += cd->name_entry_size;
6975 }
6976
6977 /* Found the named subpattern. If the name is duplicated, add one to
6978 the opcode to change CREF/RREF into DNCREF/DNRREF and insert
6979 appropriate data values. Otherwise, just insert the unique subpattern
6980 number. */
6981
6982 if (i < cd->names_found)
6983 {
6984 int offset = i++;
6985 int count = 1;
6986 recno = GET2(slot, 0); /* Number from first found */
6987 if (recno > cd->top_backref) cd->top_backref = recno;
6988 for (; i < cd->names_found; i++)
6989 {
6990 slot += cd->name_entry_size;
6991 if (STRNCMP_UC_UC(name, slot+IMM2_SIZE, namelen) != 0 ||
6992 (slot+IMM2_SIZE)[namelen] != 0) break;
6993 count++;
6994 }
6995
6996 if (count > 1)
6997 {
6998 PUT2(code, 2+LINK_SIZE, offset);
6999 PUT2(code, 2+LINK_SIZE+IMM2_SIZE, count);
7000 skipbytes += IMM2_SIZE;
7001 code[1+LINK_SIZE]++;
7002 }
7003 else /* Not a duplicated name */
7004 {
7005 PUT2(code, 2+LINK_SIZE, recno);
7006 }
7007 }
7008
7009 /* If terminator == CHAR_NULL it means that the name followed directly
7010 after the opening parenthesis [e.g. (?(abc)...] and in this case there
7011 are some further alternatives to try. For the cases where terminator !=
7012 CHAR_NULL [things like (?(<name>... or (?('name')... or (?(R&name)... ]
7013 we have now checked all the possibilities, so give an error. */
7014
7015 else if (terminator != CHAR_NULL)
7016 {
7017 *errorcodeptr = ERR15;
7018 goto FAILED;
7019 }
7020
7021 /* Check for (?(R) for recursion. Allow digits after R to specify a
7022 specific group number. */
7023
7024 else if (*name == CHAR_R)
7025 {
7026 recno = 0;
7027 for (i = 1; i < namelen; i++)
7028 {
7029 if (!IS_DIGIT(name[i]))
7030 {
7031 *errorcodeptr = ERR15;
7032 goto FAILED;
7033 }
7034 if (recno > INT_MAX / 10 - 1) /* Integer overflow */
7035 {
7036 *errorcodeptr = ERR61;
7037 goto FAILED;
7038 }
7039 recno = recno * 10 + name[i] - CHAR_0;
7040 }
7041 if (recno == 0) recno = RREF_ANY;
7042 code[1+LINK_SIZE] = OP_RREF; /* Change test type */
7043 PUT2(code, 2+LINK_SIZE, recno);
7044 }
7045
7046 /* Similarly, check for the (?(DEFINE) "condition", which is always
7047 false. */
7048
7049 else if (namelen == 6 && STRNCMP_UC_C8(name, STRING_DEFINE, 6) == 0)
7050 {
7051 code[1+LINK_SIZE] = OP_DEF;
7052 skipbytes = 1;
7053 }
7054
7055 /* Reference to an unidentified subpattern. */
7056
7057 else
7058 {
7059 *errorcodeptr = ERR15;
7060 goto FAILED;
7061 }
7062 break;
7063
7064
7065 /* ------------------------------------------------------------ */
7066 case CHAR_EQUALS_SIGN: /* Positive lookahead */
7067 bravalue = OP_ASSERT;
7068 cd->assert_depth += 1;
7069 ptr++;
7070 break;
7071
7072 /* Optimize (?!) to (*FAIL) unless it is quantified - which is a weird
7073 thing to do, but Perl allows all assertions to be quantified, and when
7074 they contain capturing parentheses there may be a potential use for
7075 this feature. Not that that applies to a quantified (?!) but we allow
7076 it for uniformity. */
7077
7078 /* ------------------------------------------------------------ */
7079 case CHAR_EXCLAMATION_MARK: /* Negative lookahead */
7080 ptr++;
7081 if (*ptr == CHAR_RIGHT_PARENTHESIS && ptr[1] != CHAR_ASTERISK &&
7082 ptr[1] != CHAR_PLUS && ptr[1] != CHAR_QUESTION_MARK &&
7083 (ptr[1] != CHAR_LEFT_CURLY_BRACKET || !is_counted_repeat(ptr+2)))
7084 {
7085 *code++ = OP_FAIL;
7086 previous = NULL;
7087 continue;
7088 }
7089 bravalue = OP_ASSERT_NOT;
7090 cd->assert_depth += 1;
7091 break;
7092
7093
7094 /* ------------------------------------------------------------ */
7095 case CHAR_LESS_THAN_SIGN: /* Lookbehind or named define */
7096 switch (ptr[1])
7097 {
7098 case CHAR_EQUALS_SIGN: /* Positive lookbehind */
7099 bravalue = OP_ASSERTBACK;
7100 cd->assert_depth += 1;
7101 ptr += 2;
7102 break;
7103
7104 case CHAR_EXCLAMATION_MARK: /* Negative lookbehind */
7105 bravalue = OP_ASSERTBACK_NOT;
7106 cd->assert_depth += 1;
7107 ptr += 2;
7108 break;
7109
7110 default: /* Could be name define, else bad */
7111 if (MAX_255(ptr[1]) && (cd->ctypes[ptr[1]] & ctype_word) != 0)
7112 goto DEFINE_NAME;
7113 ptr++; /* Correct offset for error */
7114 *errorcodeptr = ERR24;
7115 goto FAILED;
7116 }
7117 break;
7118
7119
7120 /* ------------------------------------------------------------ */
7121 case CHAR_GREATER_THAN_SIGN: /* One-time brackets */
7122 bravalue = OP_ONCE;
7123 ptr++;
7124 break;
7125
7126
7127 /* ------------------------------------------------------------ */
7128 case CHAR_C: /* Callout - may be followed by digits; */
7129 previous_callout = code; /* Save for later completion */
7130 after_manual_callout = 1; /* Skip one item before completing */
7131 *code++ = OP_CALLOUT;
7132 {
7133 int n = 0;
7134 ptr++;
7135 while(IS_DIGIT(*ptr))
7136 {
7137 n = n * 10 + *ptr++ - CHAR_0;
7138 if (n > 255)
7139 {
7140 *errorcodeptr = ERR38;
7141 goto FAILED;
7142 }
7143 }
7144 if (*ptr != CHAR_RIGHT_PARENTHESIS)
7145 {
7146 *errorcodeptr = ERR39;
7147 goto FAILED;
7148 }
7149 *code++ = n;
7150 PUT(code, 0, (int)(ptr - cd->start_pattern + 1)); /* Pattern offset */
7151 PUT(code, LINK_SIZE, 0); /* Default length */
7152 code += 2 * LINK_SIZE;
7153 }
7154 previous = NULL;
7155 continue;
7156
7157
7158 /* ------------------------------------------------------------ */
7159 case CHAR_P: /* Python-style named subpattern handling */
7160 if (*(++ptr) == CHAR_EQUALS_SIGN ||
7161 *ptr == CHAR_GREATER_THAN_SIGN) /* Reference or recursion */
7162 {
7163 is_recurse = *ptr == CHAR_GREATER_THAN_SIGN;
7164 terminator = CHAR_RIGHT_PARENTHESIS;
7165 goto NAMED_REF_OR_RECURSE;
7166 }
7167 else if (*ptr != CHAR_LESS_THAN_SIGN) /* Test for Python-style defn */
7168 {
7169 *errorcodeptr = ERR41;
7170 goto FAILED;
7171 }
7172 /* Fall through to handle (?P< as (?< is handled */
7173 /* fall through */
7174
7175 /* ------------------------------------------------------------ */
7176 DEFINE_NAME: /* Come here from (?< handling */
7177 case CHAR_APOSTROPHE:
7178 terminator = (*ptr == CHAR_LESS_THAN_SIGN)?
7179 CHAR_GREATER_THAN_SIGN : CHAR_APOSTROPHE;
7180 name = ++ptr;
7181 if (IS_DIGIT(*ptr))
7182 {
7183 *errorcodeptr = ERR84; /* Group name must start with non-digit */
7184 goto FAILED;
7185 }
7186 while (MAX_255(*ptr) && (cd->ctypes[*ptr] & ctype_word) != 0) ptr++;
7187 namelen = (int)(ptr - name);
7188
7189 /* In the pre-compile phase, do a syntax check, remember the longest
7190 name, and then remember the group in a vector, expanding it if
7191 necessary. Duplicates for the same number are skipped; other duplicates
7192 are checked for validity. In the actual compile, there is nothing to
7193 do. */
7194
7195 if (lengthptr != NULL)
7196 {
7197 named_group *ng;
7198 pcre_uint32 number = cd->bracount + 1;
7199
7200 if (*ptr != (pcre_uchar)terminator)
7201 {
7202 *errorcodeptr = ERR42;
7203 goto FAILED;
7204 }
7205
7206 if (cd->names_found >= MAX_NAME_COUNT)
7207 {
7208 *errorcodeptr = ERR49;
7209 goto FAILED;
7210 }
7211
7212 if (namelen + IMM2_SIZE + 1 > cd->name_entry_size)
7213 {
7214 cd->name_entry_size = namelen + IMM2_SIZE + 1;
7215 if (namelen > MAX_NAME_SIZE)
7216 {
7217 *errorcodeptr = ERR48;
7218 goto FAILED;
7219 }
7220 }
7221
7222 /* Scan the list to check for duplicates. For duplicate names, if the
7223 number is the same, break the loop, which causes the name to be
7224 discarded; otherwise, if DUPNAMES is not set, give an error.
7225 If it is set, allow the name with a different number, but continue
7226 scanning in case this is a duplicate with the same number. For
7227 non-duplicate names, give an error if the number is duplicated. */
7228
7229 ng = cd->named_groups;
7230 for (i = 0; i < cd->names_found; i++, ng++)
7231 {
7232 if (namelen == ng->length &&
7233 STRNCMP_UC_UC(name, ng->name, namelen) == 0)
7234 {
7235 if (ng->number == number) break;
7236 if ((options & PCRE_DUPNAMES) == 0)
7237 {
7238 *errorcodeptr = ERR43;
7239 goto FAILED;
7240 }
7241 cd->dupnames = TRUE; /* Duplicate names exist */
7242 }
7243 else if (ng->number == number)
7244 {
7245 *errorcodeptr = ERR65;
7246 goto FAILED;
7247 }
7248 }
7249
7250 if (i >= cd->names_found) /* Not a duplicate with same number */
7251 {
7252 /* Increase the list size if necessary */
7253
7254 if (cd->names_found >= cd->named_group_list_size)
7255 {
7256 int newsize = cd->named_group_list_size * 2;
7257 named_group *newspace = (PUBL(malloc))
7258 (newsize * sizeof(named_group));
7259
7260 if (newspace == NULL)
7261 {
7262 *errorcodeptr = ERR21;
7263 goto FAILED;
7264 }
7265
7266 memcpy(newspace, cd->named_groups,
7267 cd->named_group_list_size * sizeof(named_group));
7268 if (cd->named_group_list_size > NAMED_GROUP_LIST_SIZE)
7269 (PUBL(free))((void *)cd->named_groups);
7270 cd->named_groups = newspace;
7271 cd->named_group_list_size = newsize;
7272 }
7273
7274 cd->named_groups[cd->names_found].name = name;
7275 cd->named_groups[cd->names_found].length = namelen;
7276 cd->named_groups[cd->names_found].number = number;
7277 cd->names_found++;
7278 }
7279 }
7280
7281 ptr++; /* Move past > or ' in both passes. */
7282 goto NUMBERED_GROUP;
7283
7284
7285 /* ------------------------------------------------------------ */
7286 case CHAR_AMPERSAND: /* Perl recursion/subroutine syntax */
7287 terminator = CHAR_RIGHT_PARENTHESIS;
7288 is_recurse = TRUE;
7289 /* Fall through */
7290
7291 /* We come here from the Python syntax above that handles both
7292 references (?P=name) and recursion (?P>name), as well as falling
7293 through from the Perl recursion syntax (?&name). We also come here from
7294 the Perl \k<name> or \k'name' back reference syntax and the \k{name}
7295 .NET syntax, and the Oniguruma \g<...> and \g'...' subroutine syntax. */
7296
7297 NAMED_REF_OR_RECURSE:
7298 name = ++ptr;
7299 if (IS_DIGIT(*ptr))
7300 {
7301 *errorcodeptr = ERR84; /* Group name must start with non-digit */
7302 goto FAILED;
7303 }
7304 while (MAX_255(*ptr) && (cd->ctypes[*ptr] & ctype_word) != 0) ptr++;
7305 namelen = (int)(ptr - name);
7306
7307 /* In the pre-compile phase, do a syntax check. We used to just set
7308 a dummy reference number, because it was not used in the first pass.
7309 However, with the change of recursive back references to be atomic,
7310 we have to look for the number so that this state can be identified, as
7311 otherwise the incorrect length is computed. If it's not a backwards
7312 reference, the dummy number will do. */
7313
7314 if (lengthptr != NULL)
7315 {
7316 named_group *ng;
7317 recno = 0;
7318
7319 if (namelen == 0)
7320 {
7321 *errorcodeptr = ERR62;
7322 goto FAILED;
7323 }
7324 if (*ptr != (pcre_uchar)terminator)
7325 {
7326 *errorcodeptr = ERR42;
7327 goto FAILED;
7328 }
7329 if (namelen > MAX_NAME_SIZE)
7330 {
7331 *errorcodeptr = ERR48;
7332 goto FAILED;
7333 }
7334
7335 /* Count named back references. */
7336
7337 if (!is_recurse) cd->namedrefcount++;
7338
7339 /* We have to allow for a named reference to a duplicated name (this
7340 cannot be determined until the second pass). This needs an extra
7341 16-bit data item. */
7342
7343 *lengthptr += IMM2_SIZE;
7344
7345 /* If this is a forward reference and we are within a (?|...) group,
7346 the reference may end up as the number of a group which we are
7347 currently inside, that is, it could be a recursive reference. In the
7348 real compile this will be picked up and the reference wrapped with
7349 OP_ONCE to make it atomic, so we must space in case this occurs. */
7350
7351 /* In fact, this can happen for a non-forward reference because
7352 another group with the same number might be created later. This
7353 issue is fixed "properly" in PCRE2. As PCRE1 is now in maintenance
7354 only mode, we finesse the bug by allowing more memory always. */
7355
7356 *lengthptr += 4 + 4*LINK_SIZE;
7357
7358 /* It is even worse than that. The current reference may be to an
7359 existing named group with a different number (so apparently not
7360 recursive) but which later on is also attached to a group with the
7361 current number. This can only happen if $(| has been previous
7362 encountered. In that case, we allow yet more memory, just in case.
7363 (Again, this is fixed "properly" in PCRE2. */
7364
7365 if (cd->dupgroups) *lengthptr += 4 + 4*LINK_SIZE;
7366
7367 /* Otherwise, check for recursion here. The name table does not exist
7368 in the first pass; instead we must scan the list of names encountered
7369 so far in order to get the number. If the name is not found, leave
7370 the value of recno as 0 for a forward reference. */
7371
7372 /* This patch (removing "else") fixes a problem when a reference is
7373 to multiple identically named nested groups from within the nest.
7374 Once again, it is not the "proper" fix, and it results in an
7375 over-allocation of memory. */
7376
7377 /* else */
7378 {
7379 ng = cd->named_groups;
7380 for (i = 0; i < cd->names_found; i++, ng++)
7381 {
7382 if (namelen == ng->length &&
7383 STRNCMP_UC_UC(name, ng->name, namelen) == 0)
7384 {
7385 open_capitem *oc;
7386 recno = ng->number;
7387 if (is_recurse) break;
7388 for (oc = cd->open_caps; oc != NULL; oc = oc->next)
7389 {
7390 if (oc->number == recno)
7391 {
7392 oc->flag = TRUE;
7393 break;
7394 }
7395 }
7396 }
7397 }
7398 }
7399 }
7400
7401 /* In the real compile, search the name table. We check the name
7402 first, and then check that we have reached the end of the name in the
7403 table. That way, if the name is longer than any in the table, the
7404 comparison will fail without reading beyond the table entry. */
7405
7406 else
7407 {
7408 slot = cd->name_table;
7409 for (i = 0; i < cd->names_found; i++)
7410 {
7411 if (STRNCMP_UC_UC(name, slot+IMM2_SIZE, namelen) == 0 &&
7412 slot[IMM2_SIZE+namelen] == 0)
7413 break;
7414 slot += cd->name_entry_size;
7415 }
7416
7417 if (i < cd->names_found)
7418 {
7419 recno = GET2(slot, 0);
7420 }
7421 else
7422 {
7423 *errorcodeptr = ERR15;
7424 goto FAILED;
7425 }
7426 }
7427
7428 /* In both phases, for recursions, we can now go to the code than
7429 handles numerical recursion. */
7430
7431 if (is_recurse) goto HANDLE_RECURSION;
7432
7433 /* In the second pass we must see if the name is duplicated. If so, we
7434 generate a different opcode. */
7435
7436 if (lengthptr == NULL && cd->dupnames)
7437 {
7438 int count = 1;
7439 unsigned int index = i;
7440 pcre_uchar *cslot = slot + cd->name_entry_size;
7441
7442 for (i++; i < cd->names_found; i++)
7443 {
7444 if (STRCMP_UC_UC(slot + IMM2_SIZE, cslot + IMM2_SIZE) != 0) break;
7445 count++;
7446 cslot += cd->name_entry_size;
7447 }
7448
7449 if (count > 1)
7450 {
7451 if (firstcharflags == REQ_UNSET) firstcharflags = REQ_NONE;
7452 previous = code;
7453 item_hwm_offset = cd->hwm - cd->start_workspace;
7454 *code++ = ((options & PCRE_CASELESS) != 0)? OP_DNREFI : OP_DNREF;
7455 PUT2INC(code, 0, index);
7456 PUT2INC(code, 0, count);
7457
7458 /* Process each potentially referenced group. */
7459
7460 for (; slot < cslot; slot += cd->name_entry_size)
7461 {
7462 open_capitem *oc;
7463 recno = GET2(slot, 0);
7464 cd->backref_map |= (recno < 32)? (1U << recno) : 1;
7465 if (recno > cd->top_backref) cd->top_backref = recno;
7466
7467 /* Check to see if this back reference is recursive, that it, it
7468 is inside the group that it references. A flag is set so that the
7469 group can be made atomic. */
7470
7471 for (oc = cd->open_caps; oc != NULL; oc = oc->next)
7472 {
7473 if (oc->number == recno)
7474 {
7475 oc->flag = TRUE;
7476 break;
7477 }
7478 }
7479 }
7480
7481 continue; /* End of back ref handling */
7482 }
7483 }
7484
7485 /* First pass, or a non-duplicated name. */
7486
7487 goto HANDLE_REFERENCE;
7488
7489
7490 /* ------------------------------------------------------------ */
7491 case CHAR_R: /* Recursion, same as (?0) */
7492 recno = 0;
7493 if (*(++ptr) != CHAR_RIGHT_PARENTHESIS)
7494 {
7495 *errorcodeptr = ERR29;
7496 goto FAILED;
7497 }
7498 goto HANDLE_RECURSION;
7499
7500
7501 /* ------------------------------------------------------------ */
7502 case CHAR_MINUS: case CHAR_PLUS: /* Recursion or subroutine */
7503 case CHAR_0: case CHAR_1: case CHAR_2: case CHAR_3: case CHAR_4:
7504 case CHAR_5: case CHAR_6: case CHAR_7: case CHAR_8: case CHAR_9:
7505 {
7506 const pcre_uchar *called;
7507 terminator = CHAR_RIGHT_PARENTHESIS;
7508
7509 /* Come here from the \g<...> and \g'...' code (Oniguruma
7510 compatibility). However, the syntax has been checked to ensure that
7511 the ... are a (signed) number, so that neither ERR63 nor ERR29 will
7512 be called on this path, nor with the jump to OTHER_CHAR_AFTER_QUERY
7513 ever be taken. */
7514
7515 HANDLE_NUMERICAL_RECURSION:
7516
7517 if ((refsign = *ptr) == CHAR_PLUS)
7518 {
7519 ptr++;
7520 if (!IS_DIGIT(*ptr))
7521 {
7522 *errorcodeptr = ERR63;
7523 goto FAILED;
7524 }
7525 }
7526 else if (refsign == CHAR_MINUS)
7527 {
7528 if (!IS_DIGIT(ptr[1]))
7529 goto OTHER_CHAR_AFTER_QUERY;
7530 ptr++;
7531 }
7532
7533 recno = 0;
7534 while(IS_DIGIT(*ptr))
7535 {
7536 if (recno > INT_MAX / 10 - 1) /* Integer overflow */
7537 {
7538 while (IS_DIGIT(*ptr)) ptr++;
7539 *errorcodeptr = ERR61;
7540 goto FAILED;
7541 }
7542 recno = recno * 10 + *ptr++ - CHAR_0;
7543 }
7544
7545 if (*ptr != (pcre_uchar)terminator)
7546 {
7547 *errorcodeptr = ERR29;
7548 goto FAILED;
7549 }
7550
7551 if (refsign == CHAR_MINUS)
7552 {
7553 if (recno == 0)
7554 {
7555 *errorcodeptr = ERR58;
7556 goto FAILED;
7557 }
7558 recno = cd->bracount - recno + 1;
7559 if (recno <= 0)
7560 {
7561 *errorcodeptr = ERR15;
7562 goto FAILED;
7563 }
7564 }
7565 else if (refsign == CHAR_PLUS)
7566 {
7567 if (recno == 0)
7568 {
7569 *errorcodeptr = ERR58;
7570 goto FAILED;
7571 }
7572 recno += cd->bracount;
7573 }
7574
7575 /* Come here from code above that handles a named recursion */
7576
7577 HANDLE_RECURSION:
7578
7579 previous = code;
7580 item_hwm_offset = cd->hwm - cd->start_workspace;
7581 called = cd->start_code;
7582
7583 /* When we are actually compiling, find the bracket that is being
7584 referenced. Temporarily end the regex in case it doesn't exist before
7585 this point. If we end up with a forward reference, first check that
7586 the bracket does occur later so we can give the error (and position)
7587 now. Then remember this forward reference in the workspace so it can
7588 be filled in at the end. */
7589
7590 if (lengthptr == NULL)
7591 {
7592 *code = OP_END;
7593 if (recno != 0)
7594 called = PRIV(find_bracket)(cd->start_code, utf, recno);
7595
7596 /* Forward reference */
7597
7598 if (called == NULL)
7599 {
7600 if (recno > cd->final_bracount)
7601 {
7602 *errorcodeptr = ERR15;
7603 goto FAILED;
7604 }
7605
7606 /* Fudge the value of "called" so that when it is inserted as an
7607 offset below, what it actually inserted is the reference number
7608 of the group. Then remember the forward reference. */
7609
7610 called = cd->start_code + recno;
7611 if (cd->hwm >= cd->start_workspace + cd->workspace_size -
7612 WORK_SIZE_SAFETY_MARGIN)
7613 {
7614 *errorcodeptr = expand_workspace(cd);
7615 if (*errorcodeptr != 0) goto FAILED;
7616 }
7617 PUTINC(cd->hwm, 0, (int)(code + 1 - cd->start_code));
7618 }
7619
7620 /* If not a forward reference, and the subpattern is still open,
7621 this is a recursive call. We check to see if this is a left
7622 recursion that could loop for ever, and diagnose that case. We
7623 must not, however, do this check if we are in a conditional
7624 subpattern because the condition might be testing for recursion in
7625 a pattern such as /(?(R)a+|(?R)b)/, which is perfectly valid.
7626 Forever loops are also detected at runtime, so those that occur in
7627 conditional subpatterns will be picked up then. */
7628
7629 else if (GET(called, 1) == 0 && cond_depth <= 0 &&
7630 could_be_empty(called, code, bcptr, utf, cd))
7631 {
7632 *errorcodeptr = ERR40;
7633 goto FAILED;
7634 }
7635 }
7636
7637 /* Insert the recursion/subroutine item. It does not have a set first
7638 character (relevant if it is repeated, because it will then be
7639 wrapped with ONCE brackets). */
7640
7641 *code = OP_RECURSE;
7642 PUT(code, 1, (int)(called - cd->start_code));
7643 code += 1 + LINK_SIZE;
7644 groupsetfirstchar = FALSE;
7645 }
7646
7647 /* Can't determine a first byte now */
7648
7649 if (firstcharflags == REQ_UNSET) firstcharflags = REQ_NONE;
7650 zerofirstchar = firstchar;
7651 zerofirstcharflags = firstcharflags;
7652 continue;
7653
7654
7655 /* ------------------------------------------------------------ */
7656 default: /* Other characters: check option setting */
7657 OTHER_CHAR_AFTER_QUERY:
7658 set = unset = 0;
7659 optset = &set;
7660
7661 while (*ptr != CHAR_RIGHT_PARENTHESIS && *ptr != CHAR_COLON)
7662 {
7663 switch (*ptr++)
7664 {
7665 case CHAR_MINUS: optset = &unset; break;
7666
7667 case CHAR_J: /* Record that it changed in the external options */
7668 *optset |= PCRE_DUPNAMES;
7669 cd->external_flags |= PCRE_JCHANGED;
7670 break;
7671
7672 case CHAR_i: *optset |= PCRE_CASELESS; break;
7673 case CHAR_m: *optset |= PCRE_MULTILINE; break;
7674 case CHAR_s: *optset |= PCRE_DOTALL; break;
7675 case CHAR_x: *optset |= PCRE_EXTENDED; break;
7676 case CHAR_U: *optset |= PCRE_UNGREEDY; break;
7677 case CHAR_X: *optset |= PCRE_EXTRA; break;
7678
7679 default: *errorcodeptr = ERR12;
7680 ptr--; /* Correct the offset */
7681 goto FAILED;
7682 }
7683 }
7684
7685 /* Set up the changed option bits, but don't change anything yet. */
7686
7687 newoptions = (options | set) & (~unset);
7688
7689 /* If the options ended with ')' this is not the start of a nested
7690 group with option changes, so the options change at this level.
7691 If we are not at the pattern start, reset the greedy defaults and the
7692 case value for firstchar and reqchar. */
7693
7694 if (*ptr == CHAR_RIGHT_PARENTHESIS)
7695 {
7696 greedy_default = ((newoptions & PCRE_UNGREEDY) != 0);
7697 greedy_non_default = greedy_default ^ 1;
7698 req_caseopt = ((newoptions & PCRE_CASELESS) != 0)? REQ_CASELESS:0;
7699
7700 /* Change options at this level, and pass them back for use
7701 in subsequent branches. */
7702
7703 *optionsptr = options = newoptions;
7704 previous = NULL; /* This item can't be repeated */
7705 continue; /* It is complete */
7706 }
7707
7708 /* If the options ended with ':' we are heading into a nested group
7709 with possible change of options. Such groups are non-capturing and are
7710 not assertions of any kind. All we need to do is skip over the ':';
7711 the newoptions value is handled below. */
7712
7713 bravalue = OP_BRA;
7714 ptr++;
7715 } /* End of switch for character following (? */
7716 } /* End of (? handling */
7717
7718 /* Opening parenthesis not followed by '*' or '?'. If PCRE_NO_AUTO_CAPTURE
7719 is set, all unadorned brackets become non-capturing and behave like (?:...)
7720 brackets. */
7721
7722 else if ((options & PCRE_NO_AUTO_CAPTURE) != 0)
7723 {
7724 bravalue = OP_BRA;
7725 }
7726
7727 /* Else we have a capturing group. */
7728
7729 else
7730 {
7731 NUMBERED_GROUP:
7732 cd->bracount += 1;
7733 PUT2(code, 1+LINK_SIZE, cd->bracount);
7734 skipbytes = IMM2_SIZE;
7735 }
7736
7737 /* Process nested bracketed regex. First check for parentheses nested too
7738 deeply. */
7739
7740 if ((cd->parens_depth += 1) > PARENS_NEST_LIMIT)
7741 {
7742 *errorcodeptr = ERR82;
7743 goto FAILED;
7744 }
7745
7746 /* All assertions used not to be repeatable, but this was changed for Perl
7747 compatibility. All kinds can now be repeated except for assertions that are
7748 conditions (Perl also forbids these to be repeated). We copy code into a
7749 non-register variable (tempcode) in order to be able to pass its address
7750 because some compilers complain otherwise. At the start of a conditional
7751 group whose condition is an assertion, cd->iscondassert is set. We unset it
7752 here so as to allow assertions later in the group to be quantified. */
7753
7754 if (bravalue >= OP_ASSERT && bravalue <= OP_ASSERTBACK_NOT &&
7755 cd->iscondassert)
7756 {
7757 previous = NULL;
7758 cd->iscondassert = FALSE;
7759 }
7760 else
7761 {
7762 previous = code;
7763 item_hwm_offset = cd->hwm - cd->start_workspace;
7764 }
7765
7766 *code = bravalue;
7767 tempcode = code;
7768 tempreqvary = cd->req_varyopt; /* Save value before bracket */
7769 tempbracount = cd->bracount; /* Save value before bracket */
7770 length_prevgroup = 0; /* Initialize for pre-compile phase */
7771
7772 if (!compile_regex(
7773 newoptions, /* The complete new option state */
7774 &tempcode, /* Where to put code (updated) */
7775 &ptr, /* Input pointer (updated) */
7776 errorcodeptr, /* Where to put an error message */
7777 (bravalue == OP_ASSERTBACK ||
7778 bravalue == OP_ASSERTBACK_NOT), /* TRUE if back assert */
7779 reset_bracount, /* True if (?| group */
7780 skipbytes, /* Skip over bracket number */
7781 cond_depth +
7782 ((bravalue == OP_COND)?1:0), /* Depth of condition subpatterns */
7783 &subfirstchar, /* For possible first char */
7784 &subfirstcharflags,
7785 &subreqchar, /* For possible last char */
7786 &subreqcharflags,
7787 bcptr, /* Current branch chain */
7788 cd, /* Tables block */
7789 (lengthptr == NULL)? NULL : /* Actual compile phase */
7790 &length_prevgroup /* Pre-compile phase */
7791 ))
7792 goto FAILED;
7793
7794 cd->parens_depth -= 1;
7795
7796 /* If this was an atomic group and there are no capturing groups within it,
7797 generate OP_ONCE_NC instead of OP_ONCE. */
7798
7799 if (bravalue == OP_ONCE && cd->bracount <= tempbracount)
7800 *code = OP_ONCE_NC;
7801
7802 if (bravalue >= OP_ASSERT && bravalue <= OP_ASSERTBACK_NOT)
7803 cd->assert_depth -= 1;
7804
7805 /* At the end of compiling, code is still pointing to the start of the
7806 group, while tempcode has been updated to point past the end of the group.
7807 The pattern pointer (ptr) is on the bracket.
7808
7809 If this is a conditional bracket, check that there are no more than
7810 two branches in the group, or just one if it's a DEFINE group. We do this
7811 in the real compile phase, not in the pre-pass, where the whole group may
7812 not be available. */
7813
7814 if (bravalue == OP_COND && lengthptr == NULL)
7815 {
7816 pcre_uchar *tc = code;
7817 int condcount = 0;
7818
7819 do {
7820 condcount++;
7821 tc += GET(tc,1);
7822 }
7823 while (*tc != OP_KET);
7824
7825 /* A DEFINE group is never obeyed inline (the "condition" is always
7826 false). It must have only one branch. */
7827
7828 if (code[LINK_SIZE+1] == OP_DEF)
7829 {
7830 if (condcount > 1)
7831 {
7832 *errorcodeptr = ERR54;
7833 goto FAILED;
7834 }
7835 bravalue = OP_DEF; /* Just a flag to suppress char handling below */
7836 }
7837
7838 /* A "normal" conditional group. If there is just one branch, we must not
7839 make use of its firstchar or reqchar, because this is equivalent to an
7840 empty second branch. */
7841
7842 else
7843 {
7844 if (condcount > 2)
7845 {
7846 *errorcodeptr = ERR27;
7847 goto FAILED;
7848 }
7849 if (condcount == 1) subfirstcharflags = subreqcharflags = REQ_NONE;
7850 }
7851 }
7852
7853 /* Error if hit end of pattern */
7854
7855 if (*ptr != CHAR_RIGHT_PARENTHESIS)
7856 {
7857 *errorcodeptr = ERR14;
7858 goto FAILED;
7859 }
7860
7861 /* In the pre-compile phase, update the length by the length of the group,
7862 less the brackets at either end. Then reduce the compiled code to just a
7863 set of non-capturing brackets so that it doesn't use much memory if it is
7864 duplicated by a quantifier.*/
7865
7866 if (lengthptr != NULL)
7867 {
7868 if (OFLOW_MAX - *lengthptr < length_prevgroup - 2 - 2*LINK_SIZE)
7869 {
7870 *errorcodeptr = ERR20;
7871 goto FAILED;
7872 }
7873 *lengthptr += length_prevgroup - 2 - 2*LINK_SIZE;
7874 code++; /* This already contains bravalue */
7875 PUTINC(code, 0, 1 + LINK_SIZE);
7876 *code++ = OP_KET;
7877 PUTINC(code, 0, 1 + LINK_SIZE);
7878 break; /* No need to waste time with special character handling */
7879 }
7880
7881 /* Otherwise update the main code pointer to the end of the group. */
7882
7883 code = tempcode;
7884
7885 /* For a DEFINE group, required and first character settings are not
7886 relevant. */
7887
7888 if (bravalue == OP_DEF) break;
7889
7890 /* Handle updating of the required and first characters for other types of
7891 group. Update for normal brackets of all kinds, and conditions with two
7892 branches (see code above). If the bracket is followed by a quantifier with
7893 zero repeat, we have to back off. Hence the definition of zeroreqchar and
7894 zerofirstchar outside the main loop so that they can be accessed for the
7895 back off. */
7896
7897 zeroreqchar = reqchar;
7898 zeroreqcharflags = reqcharflags;
7899 zerofirstchar = firstchar;
7900 zerofirstcharflags = firstcharflags;
7901 groupsetfirstchar = FALSE;
7902
7903 if (bravalue >= OP_ONCE)
7904 {
7905 /* If we have not yet set a firstchar in this branch, take it from the
7906 subpattern, remembering that it was set here so that a repeat of more
7907 than one can replicate it as reqchar if necessary. If the subpattern has
7908 no firstchar, set "none" for the whole branch. In both cases, a zero
7909 repeat forces firstchar to "none". */
7910
7911 if (firstcharflags == REQ_UNSET)
7912 {
7913 if (subfirstcharflags >= 0)
7914 {
7915 firstchar = subfirstchar;
7916 firstcharflags = subfirstcharflags;
7917 groupsetfirstchar = TRUE;
7918 }
7919 else firstcharflags = REQ_NONE;
7920 zerofirstcharflags = REQ_NONE;
7921 }
7922
7923 /* If firstchar was previously set, convert the subpattern's firstchar
7924 into reqchar if there wasn't one, using the vary flag that was in
7925 existence beforehand. */
7926
7927 else if (subfirstcharflags >= 0 && subreqcharflags < 0)
7928 {
7929 subreqchar = subfirstchar;
7930 subreqcharflags = subfirstcharflags | tempreqvary;
7931 }
7932
7933 /* If the subpattern set a required byte (or set a first byte that isn't
7934 really the first byte - see above), set it. */
7935
7936 if (subreqcharflags >= 0)
7937 {
7938 reqchar = subreqchar;
7939 reqcharflags = subreqcharflags;
7940 }
7941 }
7942
7943 /* For a forward assertion, we take the reqchar, if set, provided that the
7944 group has also set a first char. This can be helpful if the pattern that
7945 follows the assertion doesn't set a different char. For example, it's
7946 useful for /(?=abcde).+/. We can't set firstchar for an assertion, however
7947 because it leads to incorrect effect for patterns such as /(?=a)a.+/ when
7948 the "real" "a" would then become a reqchar instead of a firstchar. This is
7949 overcome by a scan at the end if there's no firstchar, looking for an
7950 asserted first char. */
7951
7952 else if (bravalue == OP_ASSERT && subreqcharflags >= 0 &&
7953 subfirstcharflags >= 0)
7954 {
7955 reqchar = subreqchar;
7956 reqcharflags = subreqcharflags;
7957 }
7958 break; /* End of processing '(' */
7959
7960
7961 /* ===================================================================*/
7962 /* Handle metasequences introduced by \. For ones like \d, the ESC_ values
7963 are arranged to be the negation of the corresponding OP_values in the
7964 default case when PCRE_UCP is not set. For the back references, the values
7965 are negative the reference number. Only back references and those types
7966 that consume a character may be repeated. We can test for values between
7967 ESC_b and ESC_Z for the latter; this may have to change if any new ones are
7968 ever created. */
7969
7970 case CHAR_BACKSLASH:
7971 tempptr = ptr;
7972 escape = check_escape(&ptr, &ec, errorcodeptr, cd->bracount, options, FALSE);
7973 if (*errorcodeptr != 0) goto FAILED;
7974
7975 if (escape == 0) /* The escape coded a single character */
7976 c = ec;
7977 else
7978 {
7979 /* For metasequences that actually match a character, we disable the
7980 setting of a first character if it hasn't already been set. */
7981
7982 if (firstcharflags == REQ_UNSET && escape > ESC_b && escape < ESC_Z)
7983 firstcharflags = REQ_NONE;
7984
7985 /* Set values to reset to if this is followed by a zero repeat. */
7986
7987 zerofirstchar = firstchar;
7988 zerofirstcharflags = firstcharflags;
7989 zeroreqchar = reqchar;
7990 zeroreqcharflags = reqcharflags;
7991
7992 /* \g<name> or \g'name' is a subroutine call by name and \g<n> or \g'n'
7993 is a subroutine call by number (Oniguruma syntax). In fact, the value
7994 ESC_g is returned only for these cases. So we don't need to check for <
7995 or ' if the value is ESC_g. For the Perl syntax \g{n} the value is
7996 -n, and for the Perl syntax \g{name} the result is ESC_k (as
7997 that is a synonym for a named back reference). */
7998
7999 if (escape == ESC_g)
8000 {
8001 const pcre_uchar *p;
8002 pcre_uint32 cf;
8003
8004 item_hwm_offset = cd->hwm - cd->start_workspace; /* Normally this is set when '(' is read */
8005 terminator = (*(++ptr) == CHAR_LESS_THAN_SIGN)?
8006 CHAR_GREATER_THAN_SIGN : CHAR_APOSTROPHE;
8007
8008 /* These two statements stop the compiler for warning about possibly
8009 unset variables caused by the jump to HANDLE_NUMERICAL_RECURSION. In
8010 fact, because we do the check for a number below, the paths that
8011 would actually be in error are never taken. */
8012
8013 skipbytes = 0;
8014 reset_bracount = FALSE;
8015
8016 /* If it's not a signed or unsigned number, treat it as a name. */
8017
8018 cf = ptr[1];
8019 if (cf != CHAR_PLUS && cf != CHAR_MINUS && !IS_DIGIT(cf))
8020 {
8021 is_recurse = TRUE;
8022 goto NAMED_REF_OR_RECURSE;
8023 }
8024
8025 /* Signed or unsigned number (cf = ptr[1]) is known to be plus or minus
8026 or a digit. */
8027
8028 p = ptr + 2;
8029 while (IS_DIGIT(*p)) p++;
8030 if (*p != (pcre_uchar)terminator)
8031 {
8032 *errorcodeptr = ERR57;
8033 goto FAILED;
8034 }
8035 ptr++;
8036 goto HANDLE_NUMERICAL_RECURSION;
8037 }
8038
8039 /* \k<name> or \k'name' is a back reference by name (Perl syntax).
8040 We also support \k{name} (.NET syntax). */
8041
8042 if (escape == ESC_k)
8043 {
8044 if ((ptr[1] != CHAR_LESS_THAN_SIGN &&
8045 ptr[1] != CHAR_APOSTROPHE && ptr[1] != CHAR_LEFT_CURLY_BRACKET))
8046 {
8047 *errorcodeptr = ERR69;
8048 goto FAILED;
8049 }
8050 is_recurse = FALSE;
8051 terminator = (*(++ptr) == CHAR_LESS_THAN_SIGN)?
8052 CHAR_GREATER_THAN_SIGN : (*ptr == CHAR_APOSTROPHE)?
8053 CHAR_APOSTROPHE : CHAR_RIGHT_CURLY_BRACKET;
8054 goto NAMED_REF_OR_RECURSE;
8055 }
8056
8057 /* Back references are handled specially; must disable firstchar if
8058 not set to cope with cases like (?=(\w+))\1: which would otherwise set
8059 ':' later. */
8060
8061 if (escape < 0)
8062 {
8063 open_capitem *oc;
8064 recno = -escape;
8065
8066 /* Come here from named backref handling when the reference is to a
8067 single group (i.e. not to a duplicated name. */
8068
8069 HANDLE_REFERENCE:
8070 if (firstcharflags == REQ_UNSET) zerofirstcharflags = firstcharflags = REQ_NONE;
8071 previous = code;
8072 item_hwm_offset = cd->hwm - cd->start_workspace;
8073 *code++ = ((options & PCRE_CASELESS) != 0)? OP_REFI : OP_REF;
8074 PUT2INC(code, 0, recno);
8075 cd->backref_map |= (recno < 32)? (1U << recno) : 1;
8076 if (recno > cd->top_backref) cd->top_backref = recno;
8077
8078 /* Check to see if this back reference is recursive, that it, it
8079 is inside the group that it references. A flag is set so that the
8080 group can be made atomic. */
8081
8082 for (oc = cd->open_caps; oc != NULL; oc = oc->next)
8083 {
8084 if (oc->number == recno)
8085 {
8086 oc->flag = TRUE;
8087 break;
8088 }
8089 }
8090 }
8091
8092 /* So are Unicode property matches, if supported. */
8093
8094 #ifdef SUPPORT_UCP
8095 else if (escape == ESC_P || escape == ESC_p)
8096 {
8097 BOOL negated;
8098 unsigned int ptype = 0, pdata = 0;
8099 if (!get_ucp(&ptr, &negated, &ptype, &pdata, errorcodeptr))
8100 goto FAILED;
8101 previous = code;
8102 item_hwm_offset = cd->hwm - cd->start_workspace;
8103 *code++ = ((escape == ESC_p) != negated)? OP_PROP : OP_NOTPROP;
8104 *code++ = ptype;
8105 *code++ = pdata;
8106 }
8107 #else
8108
8109 /* If Unicode properties are not supported, \X, \P, and \p are not
8110 allowed. */
8111
8112 else if (escape == ESC_X || escape == ESC_P || escape == ESC_p)
8113 {
8114 *errorcodeptr = ERR45;
8115 goto FAILED;
8116 }
8117 #endif
8118
8119 /* For the rest (including \X when Unicode properties are supported), we
8120 can obtain the OP value by negating the escape value in the default
8121 situation when PCRE_UCP is not set. When it *is* set, we substitute
8122 Unicode property tests. Note that \b and \B do a one-character
8123 lookbehind, and \A also behaves as if it does. */
8124
8125 else
8126 {
8127 if ((escape == ESC_b || escape == ESC_B || escape == ESC_A) &&
8128 cd->max_lookbehind == 0)
8129 cd->max_lookbehind = 1;
8130 #ifdef SUPPORT_UCP
8131 if (escape >= ESC_DU && escape <= ESC_wu)
8132 {
8133 nestptr = ptr + 1; /* Where to resume */
8134 ptr = substitutes[escape - ESC_DU] - 1; /* Just before substitute */
8135 }
8136 else
8137 #endif
8138 /* In non-UTF-8 mode, we turn \C into OP_ALLANY instead of OP_ANYBYTE
8139 so that it works in DFA mode and in lookbehinds. */
8140
8141 {
8142 previous = (escape > ESC_b && escape < ESC_Z)? code : NULL;
8143 item_hwm_offset = cd->hwm - cd->start_workspace;
8144 *code++ = (!utf && escape == ESC_C)? OP_ALLANY : escape;
8145 }
8146 }
8147 continue;
8148 }
8149
8150 /* We have a data character whose value is in c. In UTF-8 mode it may have
8151 a value > 127. We set its representation in the length/buffer, and then
8152 handle it as a data character. */
8153
8154 #if defined SUPPORT_UTF && !defined COMPILE_PCRE32
8155 if (utf && c > MAX_VALUE_FOR_SINGLE_CHAR)
8156 mclength = PRIV(ord2utf)(c, mcbuffer);
8157 else
8158 #endif
8159
8160 {
8161 mcbuffer[0] = c;
8162 mclength = 1;
8163 }
8164 goto ONE_CHAR;
8165
8166
8167 /* ===================================================================*/
8168 /* Handle a literal character. It is guaranteed not to be whitespace or #
8169 when the extended flag is set. If we are in a UTF mode, it may be a
8170 multi-unit literal character. */
8171
8172 default:
8173 NORMAL_CHAR:
8174 mclength = 1;
8175 mcbuffer[0] = c;
8176
8177 #ifdef SUPPORT_UTF
8178 if (utf && HAS_EXTRALEN(c))
8179 ACROSSCHAR(TRUE, ptr[1], mcbuffer[mclength++] = *(++ptr));
8180 #endif
8181
8182 /* At this point we have the character's bytes in mcbuffer, and the length
8183 in mclength. When not in UTF-8 mode, the length is always 1. */
8184
8185 ONE_CHAR:
8186 previous = code;
8187 item_hwm_offset = cd->hwm - cd->start_workspace;
8188
8189 /* For caseless UTF-8 mode when UCP support is available, check whether
8190 this character has more than one other case. If so, generate a special
8191 OP_PROP item instead of OP_CHARI. */
8192
8193 #ifdef SUPPORT_UCP
8194 if (utf && (options & PCRE_CASELESS) != 0)
8195 {
8196 GETCHAR(c, mcbuffer);
8197 if ((c = UCD_CASESET(c)) != 0)
8198 {
8199 *code++ = OP_PROP;
8200 *code++ = PT_CLIST;
8201 *code++ = c;
8202 if (firstcharflags == REQ_UNSET)
8203 firstcharflags = zerofirstcharflags = REQ_NONE;
8204 break;
8205 }
8206 }
8207 #endif
8208
8209 /* Caseful matches, or not one of the multicase characters. */
8210
8211 *code++ = ((options & PCRE_CASELESS) != 0)? OP_CHARI : OP_CHAR;
8212 for (c = 0; c < mclength; c++) *code++ = mcbuffer[c];
8213
8214 /* Remember if \r or \n were seen */
8215
8216 if (mcbuffer[0] == CHAR_CR || mcbuffer[0] == CHAR_NL)
8217 cd->external_flags |= PCRE_HASCRORLF;
8218
8219 /* Set the first and required bytes appropriately. If no previous first
8220 byte, set it from this character, but revert to none on a zero repeat.
8221 Otherwise, leave the firstchar value alone, and don't change it on a zero
8222 repeat. */
8223
8224 if (firstcharflags == REQ_UNSET)
8225 {
8226 zerofirstcharflags = REQ_NONE;
8227 zeroreqchar = reqchar;
8228 zeroreqcharflags = reqcharflags;
8229
8230 /* If the character is more than one byte long, we can set firstchar
8231 only if it is not to be matched caselessly. */
8232
8233 if (mclength == 1 || req_caseopt == 0)
8234 {
8235 firstchar = mcbuffer[0];
8236 firstcharflags = req_caseopt;
8237
8238 if (mclength != 1)
8239 {
8240 reqchar = code[-1];
8241 reqcharflags = cd->req_varyopt;
8242 }
8243 }
8244 else firstcharflags = reqcharflags = REQ_NONE;
8245 }
8246
8247 /* firstchar was previously set; we can set reqchar only if the length is
8248 1 or the matching is caseful. */
8249
8250 else
8251 {
8252 zerofirstchar = firstchar;
8253 zerofirstcharflags = firstcharflags;
8254 zeroreqchar = reqchar;
8255 zeroreqcharflags = reqcharflags;
8256 if (mclength == 1 || req_caseopt == 0)
8257 {
8258 reqchar = code[-1];
8259 reqcharflags = req_caseopt | cd->req_varyopt;
8260 }
8261 }
8262
8263 break; /* End of literal character handling */
8264 }
8265 } /* end of big loop */
8266
8267
8268 /* Control never reaches here by falling through, only by a goto for all the
8269 error states. Pass back the position in the pattern so that it can be displayed
8270 to the user for diagnosing the error. */
8271
8272 FAILED:
8273 *ptrptr = ptr;
8274 return FALSE;
8275 }
8276
8277
8278
8279 /*************************************************
8280 * Compile sequence of alternatives *
8281 *************************************************/
8282
8283 /* On entry, ptr is pointing past the bracket character, but on return it
8284 points to the closing bracket, or vertical bar, or end of string. The code
8285 variable is pointing at the byte into which the BRA operator has been stored.
8286 This function is used during the pre-compile phase when we are trying to find
8287 out the amount of memory needed, as well as during the real compile phase. The
8288 value of lengthptr distinguishes the two phases.
8289
8290 Arguments:
8291 options option bits, including any changes for this subpattern
8292 codeptr -> the address of the current code pointer
8293 ptrptr -> the address of the current pattern pointer
8294 errorcodeptr -> pointer to error code variable
8295 lookbehind TRUE if this is a lookbehind assertion
8296 reset_bracount TRUE to reset the count for each branch
8297 skipbytes skip this many bytes at start (for brackets and OP_COND)
8298 cond_depth depth of nesting for conditional subpatterns
8299 firstcharptr place to put the first required character
8300 firstcharflagsptr place to put the first character flags, or a negative number
8301 reqcharptr place to put the last required character
8302 reqcharflagsptr place to put the last required character flags, or a negative number
8303 bcptr pointer to the chain of currently open branches
8304 cd points to the data block with tables pointers etc.
8305 lengthptr NULL during the real compile phase
8306 points to length accumulator during pre-compile phase
8307
8308 Returns: TRUE on success
8309 */
8310
8311 static BOOL
compile_regex(int options,pcre_uchar ** codeptr,const pcre_uchar ** ptrptr,int * errorcodeptr,BOOL lookbehind,BOOL reset_bracount,int skipbytes,int cond_depth,pcre_uint32 * firstcharptr,pcre_int32 * firstcharflagsptr,pcre_uint32 * reqcharptr,pcre_int32 * reqcharflagsptr,branch_chain * bcptr,compile_data * cd,int * lengthptr)8312 compile_regex(int options, pcre_uchar **codeptr, const pcre_uchar **ptrptr,
8313 int *errorcodeptr, BOOL lookbehind, BOOL reset_bracount, int skipbytes,
8314 int cond_depth,
8315 pcre_uint32 *firstcharptr, pcre_int32 *firstcharflagsptr,
8316 pcre_uint32 *reqcharptr, pcre_int32 *reqcharflagsptr,
8317 branch_chain *bcptr, compile_data *cd, int *lengthptr)
8318 {
8319 const pcre_uchar *ptr = *ptrptr;
8320 pcre_uchar *code = *codeptr;
8321 pcre_uchar *last_branch = code;
8322 pcre_uchar *start_bracket = code;
8323 pcre_uchar *reverse_count = NULL;
8324 open_capitem capitem;
8325 int capnumber = 0;
8326 pcre_uint32 firstchar, reqchar;
8327 pcre_int32 firstcharflags, reqcharflags;
8328 pcre_uint32 branchfirstchar, branchreqchar;
8329 pcre_int32 branchfirstcharflags, branchreqcharflags;
8330 int length;
8331 unsigned int orig_bracount;
8332 unsigned int max_bracount;
8333 branch_chain bc;
8334 size_t save_hwm_offset;
8335
8336 /* If set, call the external function that checks for stack availability. */
8337
8338 if (PUBL(stack_guard) != NULL && PUBL(stack_guard)())
8339 {
8340 *errorcodeptr= ERR85;
8341 return FALSE;
8342 }
8343
8344 /* Miscellaneous initialization */
8345
8346 bc.outer = bcptr;
8347 bc.current_branch = code;
8348
8349 firstchar = reqchar = 0;
8350 firstcharflags = reqcharflags = REQ_UNSET;
8351
8352 save_hwm_offset = cd->hwm - cd->start_workspace;
8353
8354 /* Accumulate the length for use in the pre-compile phase. Start with the
8355 length of the BRA and KET and any extra bytes that are required at the
8356 beginning. We accumulate in a local variable to save frequent testing of
8357 lengthptr for NULL. We cannot do this by looking at the value of code at the
8358 start and end of each alternative, because compiled items are discarded during
8359 the pre-compile phase so that the work space is not exceeded. */
8360
8361 length = 2 + 2*LINK_SIZE + skipbytes;
8362
8363 /* WARNING: If the above line is changed for any reason, you must also change
8364 the code that abstracts option settings at the start of the pattern and makes
8365 them global. It tests the value of length for (2 + 2*LINK_SIZE) in the
8366 pre-compile phase to find out whether anything has yet been compiled or not. */
8367
8368 /* If this is a capturing subpattern, add to the chain of open capturing items
8369 so that we can detect them if (*ACCEPT) is encountered. This is also used to
8370 detect groups that contain recursive back references to themselves. Note that
8371 only OP_CBRA need be tested here; changing this opcode to one of its variants,
8372 e.g. OP_SCBRAPOS, happens later, after the group has been compiled. */
8373
8374 if (*code == OP_CBRA)
8375 {
8376 capnumber = GET2(code, 1 + LINK_SIZE);
8377 capitem.number = capnumber;
8378 capitem.next = cd->open_caps;
8379 capitem.flag = FALSE;
8380 cd->open_caps = &capitem;
8381 }
8382
8383 /* Offset is set zero to mark that this bracket is still open */
8384
8385 PUT(code, 1, 0);
8386 code += 1 + LINK_SIZE + skipbytes;
8387
8388 /* Loop for each alternative branch */
8389
8390 orig_bracount = max_bracount = cd->bracount;
8391 for (;;)
8392 {
8393 /* For a (?| group, reset the capturing bracket count so that each branch
8394 uses the same numbers. */
8395
8396 if (reset_bracount) cd->bracount = orig_bracount;
8397
8398 /* Set up dummy OP_REVERSE if lookbehind assertion */
8399
8400 if (lookbehind)
8401 {
8402 *code++ = OP_REVERSE;
8403 reverse_count = code;
8404 PUTINC(code, 0, 0);
8405 length += 1 + LINK_SIZE;
8406 }
8407
8408 /* Now compile the branch; in the pre-compile phase its length gets added
8409 into the length. */
8410
8411 if (!compile_branch(&options, &code, &ptr, errorcodeptr, &branchfirstchar,
8412 &branchfirstcharflags, &branchreqchar, &branchreqcharflags, &bc,
8413 cond_depth, cd, (lengthptr == NULL)? NULL : &length))
8414 {
8415 *ptrptr = ptr;
8416 return FALSE;
8417 }
8418
8419 /* Keep the highest bracket count in case (?| was used and some branch
8420 has fewer than the rest. */
8421
8422 if (cd->bracount > max_bracount) max_bracount = cd->bracount;
8423
8424 /* In the real compile phase, there is some post-processing to be done. */
8425
8426 if (lengthptr == NULL)
8427 {
8428 /* If this is the first branch, the firstchar and reqchar values for the
8429 branch become the values for the regex. */
8430
8431 if (*last_branch != OP_ALT)
8432 {
8433 firstchar = branchfirstchar;
8434 firstcharflags = branchfirstcharflags;
8435 reqchar = branchreqchar;
8436 reqcharflags = branchreqcharflags;
8437 }
8438
8439 /* If this is not the first branch, the first char and reqchar have to
8440 match the values from all the previous branches, except that if the
8441 previous value for reqchar didn't have REQ_VARY set, it can still match,
8442 and we set REQ_VARY for the regex. */
8443
8444 else
8445 {
8446 /* If we previously had a firstchar, but it doesn't match the new branch,
8447 we have to abandon the firstchar for the regex, but if there was
8448 previously no reqchar, it takes on the value of the old firstchar. */
8449
8450 if (firstcharflags >= 0 &&
8451 (firstcharflags != branchfirstcharflags || firstchar != branchfirstchar))
8452 {
8453 if (reqcharflags < 0)
8454 {
8455 reqchar = firstchar;
8456 reqcharflags = firstcharflags;
8457 }
8458 firstcharflags = REQ_NONE;
8459 }
8460
8461 /* If we (now or from before) have no firstchar, a firstchar from the
8462 branch becomes a reqchar if there isn't a branch reqchar. */
8463
8464 if (firstcharflags < 0 && branchfirstcharflags >= 0 && branchreqcharflags < 0)
8465 {
8466 branchreqchar = branchfirstchar;
8467 branchreqcharflags = branchfirstcharflags;
8468 }
8469
8470 /* Now ensure that the reqchars match */
8471
8472 if (((reqcharflags & ~REQ_VARY) != (branchreqcharflags & ~REQ_VARY)) ||
8473 reqchar != branchreqchar)
8474 reqcharflags = REQ_NONE;
8475 else
8476 {
8477 reqchar = branchreqchar;
8478 reqcharflags |= branchreqcharflags; /* To "or" REQ_VARY */
8479 }
8480 }
8481
8482 /* If lookbehind, check that this branch matches a fixed-length string, and
8483 put the length into the OP_REVERSE item. Temporarily mark the end of the
8484 branch with OP_END. If the branch contains OP_RECURSE, the result is -3
8485 because there may be forward references that we can't check here. Set a
8486 flag to cause another lookbehind check at the end. Why not do it all at the
8487 end? Because common, erroneous checks are picked up here and the offset of
8488 the problem can be shown. */
8489
8490 if (lookbehind)
8491 {
8492 int fixed_length;
8493 *code = OP_END;
8494 fixed_length = find_fixedlength(last_branch, (options & PCRE_UTF8) != 0,
8495 FALSE, cd, NULL);
8496 DPRINTF(("fixed length = %d\n", fixed_length));
8497 if (fixed_length == -3)
8498 {
8499 cd->check_lookbehind = TRUE;
8500 }
8501 else if (fixed_length < 0)
8502 {
8503 *errorcodeptr = (fixed_length == -2)? ERR36 :
8504 (fixed_length == -4)? ERR70: ERR25;
8505 *ptrptr = ptr;
8506 return FALSE;
8507 }
8508 else
8509 {
8510 if (fixed_length > cd->max_lookbehind)
8511 cd->max_lookbehind = fixed_length;
8512 PUT(reverse_count, 0, fixed_length);
8513 }
8514 }
8515 }
8516
8517 /* Reached end of expression, either ')' or end of pattern. In the real
8518 compile phase, go back through the alternative branches and reverse the chain
8519 of offsets, with the field in the BRA item now becoming an offset to the
8520 first alternative. If there are no alternatives, it points to the end of the
8521 group. The length in the terminating ket is always the length of the whole
8522 bracketed item. Return leaving the pointer at the terminating char. */
8523
8524 if (*ptr != CHAR_VERTICAL_LINE)
8525 {
8526 if (lengthptr == NULL)
8527 {
8528 int branch_length = (int)(code - last_branch);
8529 do
8530 {
8531 int prev_length = GET(last_branch, 1);
8532 PUT(last_branch, 1, branch_length);
8533 branch_length = prev_length;
8534 last_branch -= branch_length;
8535 }
8536 while (branch_length > 0);
8537 }
8538
8539 /* Fill in the ket */
8540
8541 *code = OP_KET;
8542 PUT(code, 1, (int)(code - start_bracket));
8543 code += 1 + LINK_SIZE;
8544
8545 /* If it was a capturing subpattern, check to see if it contained any
8546 recursive back references. If so, we must wrap it in atomic brackets.
8547 Because we are moving code along, we must ensure that any pending recursive
8548 references are updated. In any event, remove the block from the chain. */
8549
8550 if (capnumber > 0)
8551 {
8552 if (cd->open_caps->flag)
8553 {
8554 *code = OP_END;
8555 adjust_recurse(start_bracket, 1 + LINK_SIZE,
8556 (options & PCRE_UTF8) != 0, cd, save_hwm_offset);
8557 memmove(start_bracket + 1 + LINK_SIZE, start_bracket,
8558 IN_UCHARS(code - start_bracket));
8559 *start_bracket = OP_ONCE;
8560 code += 1 + LINK_SIZE;
8561 PUT(start_bracket, 1, (int)(code - start_bracket));
8562 *code = OP_KET;
8563 PUT(code, 1, (int)(code - start_bracket));
8564 code += 1 + LINK_SIZE;
8565 length += 2 + 2*LINK_SIZE;
8566 }
8567 cd->open_caps = cd->open_caps->next;
8568 }
8569
8570 /* Retain the highest bracket number, in case resetting was used. */
8571
8572 cd->bracount = max_bracount;
8573
8574 /* Set values to pass back */
8575
8576 *codeptr = code;
8577 *ptrptr = ptr;
8578 *firstcharptr = firstchar;
8579 *firstcharflagsptr = firstcharflags;
8580 *reqcharptr = reqchar;
8581 *reqcharflagsptr = reqcharflags;
8582 if (lengthptr != NULL)
8583 {
8584 if (OFLOW_MAX - *lengthptr < length)
8585 {
8586 *errorcodeptr = ERR20;
8587 return FALSE;
8588 }
8589 *lengthptr += length;
8590 }
8591 return TRUE;
8592 }
8593
8594 /* Another branch follows. In the pre-compile phase, we can move the code
8595 pointer back to where it was for the start of the first branch. (That is,
8596 pretend that each branch is the only one.)
8597
8598 In the real compile phase, insert an ALT node. Its length field points back
8599 to the previous branch while the bracket remains open. At the end the chain
8600 is reversed. It's done like this so that the start of the bracket has a
8601 zero offset until it is closed, making it possible to detect recursion. */
8602
8603 if (lengthptr != NULL)
8604 {
8605 code = *codeptr + 1 + LINK_SIZE + skipbytes;
8606 length += 1 + LINK_SIZE;
8607 }
8608 else
8609 {
8610 *code = OP_ALT;
8611 PUT(code, 1, (int)(code - last_branch));
8612 bc.current_branch = last_branch = code;
8613 code += 1 + LINK_SIZE;
8614 }
8615
8616 ptr++;
8617 }
8618 /* Control never reaches here */
8619 }
8620
8621
8622
8623
8624 /*************************************************
8625 * Check for anchored expression *
8626 *************************************************/
8627
8628 /* Try to find out if this is an anchored regular expression. Consider each
8629 alternative branch. If they all start with OP_SOD or OP_CIRC, or with a bracket
8630 all of whose alternatives start with OP_SOD or OP_CIRC (recurse ad lib), then
8631 it's anchored. However, if this is a multiline pattern, then only OP_SOD will
8632 be found, because ^ generates OP_CIRCM in that mode.
8633
8634 We can also consider a regex to be anchored if OP_SOM starts all its branches.
8635 This is the code for \G, which means "match at start of match position, taking
8636 into account the match offset".
8637
8638 A branch is also implicitly anchored if it starts with .* and DOTALL is set,
8639 because that will try the rest of the pattern at all possible matching points,
8640 so there is no point trying again.... er ....
8641
8642 .... except when the .* appears inside capturing parentheses, and there is a
8643 subsequent back reference to those parentheses. We haven't enough information
8644 to catch that case precisely.
8645
8646 At first, the best we could do was to detect when .* was in capturing brackets
8647 and the highest back reference was greater than or equal to that level.
8648 However, by keeping a bitmap of the first 31 back references, we can catch some
8649 of the more common cases more precisely.
8650
8651 ... A second exception is when the .* appears inside an atomic group, because
8652 this prevents the number of characters it matches from being adjusted.
8653
8654 Arguments:
8655 code points to start of expression (the bracket)
8656 bracket_map a bitmap of which brackets we are inside while testing; this
8657 handles up to substring 31; after that we just have to take
8658 the less precise approach
8659 cd points to the compile data block
8660 atomcount atomic group level
8661
8662 Returns: TRUE or FALSE
8663 */
8664
8665 static BOOL
is_anchored(register const pcre_uchar * code,unsigned int bracket_map,compile_data * cd,int atomcount)8666 is_anchored(register const pcre_uchar *code, unsigned int bracket_map,
8667 compile_data *cd, int atomcount)
8668 {
8669 do {
8670 const pcre_uchar *scode = first_significant_code(
8671 code + PRIV(OP_lengths)[*code], FALSE);
8672 register int op = *scode;
8673
8674 /* Non-capturing brackets */
8675
8676 if (op == OP_BRA || op == OP_BRAPOS ||
8677 op == OP_SBRA || op == OP_SBRAPOS)
8678 {
8679 if (!is_anchored(scode, bracket_map, cd, atomcount)) return FALSE;
8680 }
8681
8682 /* Capturing brackets */
8683
8684 else if (op == OP_CBRA || op == OP_CBRAPOS ||
8685 op == OP_SCBRA || op == OP_SCBRAPOS)
8686 {
8687 int n = GET2(scode, 1+LINK_SIZE);
8688 int new_map = bracket_map | ((n < 32)? (1U << n) : 1);
8689 if (!is_anchored(scode, new_map, cd, atomcount)) return FALSE;
8690 }
8691
8692 /* Positive forward assertion */
8693
8694 else if (op == OP_ASSERT)
8695 {
8696 if (!is_anchored(scode, bracket_map, cd, atomcount)) return FALSE;
8697 }
8698
8699 /* Condition; not anchored if no second branch */
8700
8701 else if (op == OP_COND)
8702 {
8703 if (scode[GET(scode,1)] != OP_ALT) return FALSE;
8704 if (!is_anchored(scode, bracket_map, cd, atomcount)) return FALSE;
8705 }
8706
8707 /* Atomic groups */
8708
8709 else if (op == OP_ONCE || op == OP_ONCE_NC)
8710 {
8711 if (!is_anchored(scode, bracket_map, cd, atomcount + 1))
8712 return FALSE;
8713 }
8714
8715 /* .* is not anchored unless DOTALL is set (which generates OP_ALLANY) and
8716 it isn't in brackets that are or may be referenced or inside an atomic
8717 group. */
8718
8719 else if ((op == OP_TYPESTAR || op == OP_TYPEMINSTAR ||
8720 op == OP_TYPEPOSSTAR))
8721 {
8722 if (scode[1] != OP_ALLANY || (bracket_map & cd->backref_map) != 0 ||
8723 atomcount > 0 || cd->had_pruneorskip)
8724 return FALSE;
8725 }
8726
8727 /* Check for explicit anchoring */
8728
8729 else if (op != OP_SOD && op != OP_SOM && op != OP_CIRC) return FALSE;
8730
8731 code += GET(code, 1);
8732 }
8733 while (*code == OP_ALT); /* Loop for each alternative */
8734 return TRUE;
8735 }
8736
8737
8738
8739 /*************************************************
8740 * Check for starting with ^ or .* *
8741 *************************************************/
8742
8743 /* This is called to find out if every branch starts with ^ or .* so that
8744 "first char" processing can be done to speed things up in multiline
8745 matching and for non-DOTALL patterns that start with .* (which must start at
8746 the beginning or after \n). As in the case of is_anchored() (see above), we
8747 have to take account of back references to capturing brackets that contain .*
8748 because in that case we can't make the assumption. Also, the appearance of .*
8749 inside atomic brackets or in an assertion, or in a pattern that contains *PRUNE
8750 or *SKIP does not count, because once again the assumption no longer holds.
8751
8752 Arguments:
8753 code points to start of expression (the bracket)
8754 bracket_map a bitmap of which brackets we are inside while testing; this
8755 handles up to substring 31; after that we just have to take
8756 the less precise approach
8757 cd points to the compile data
8758 atomcount atomic group level
8759 inassert TRUE if in an assertion
8760
8761 Returns: TRUE or FALSE
8762 */
8763
8764 static BOOL
is_startline(const pcre_uchar * code,unsigned int bracket_map,compile_data * cd,int atomcount,BOOL inassert)8765 is_startline(const pcre_uchar *code, unsigned int bracket_map,
8766 compile_data *cd, int atomcount, BOOL inassert)
8767 {
8768 do {
8769 const pcre_uchar *scode = first_significant_code(
8770 code + PRIV(OP_lengths)[*code], FALSE);
8771 register int op = *scode;
8772
8773 /* If we are at the start of a conditional assertion group, *both* the
8774 conditional assertion *and* what follows the condition must satisfy the test
8775 for start of line. Other kinds of condition fail. Note that there may be an
8776 auto-callout at the start of a condition. */
8777
8778 if (op == OP_COND)
8779 {
8780 scode += 1 + LINK_SIZE;
8781 if (*scode == OP_CALLOUT) scode += PRIV(OP_lengths)[OP_CALLOUT];
8782 switch (*scode)
8783 {
8784 case OP_CREF:
8785 case OP_DNCREF:
8786 case OP_RREF:
8787 case OP_DNRREF:
8788 case OP_DEF:
8789 case OP_FAIL:
8790 return FALSE;
8791
8792 default: /* Assertion */
8793 if (!is_startline(scode, bracket_map, cd, atomcount, TRUE)) return FALSE;
8794 do scode += GET(scode, 1); while (*scode == OP_ALT);
8795 scode += 1 + LINK_SIZE;
8796 break;
8797 }
8798 scode = first_significant_code(scode, FALSE);
8799 op = *scode;
8800 }
8801
8802 /* Non-capturing brackets */
8803
8804 if (op == OP_BRA || op == OP_BRAPOS ||
8805 op == OP_SBRA || op == OP_SBRAPOS)
8806 {
8807 if (!is_startline(scode, bracket_map, cd, atomcount, inassert)) return FALSE;
8808 }
8809
8810 /* Capturing brackets */
8811
8812 else if (op == OP_CBRA || op == OP_CBRAPOS ||
8813 op == OP_SCBRA || op == OP_SCBRAPOS)
8814 {
8815 int n = GET2(scode, 1+LINK_SIZE);
8816 int new_map = bracket_map | ((n < 32)? (1U << n) : 1);
8817 if (!is_startline(scode, new_map, cd, atomcount, inassert)) return FALSE;
8818 }
8819
8820 /* Positive forward assertions */
8821
8822 else if (op == OP_ASSERT)
8823 {
8824 if (!is_startline(scode, bracket_map, cd, atomcount, TRUE)) return FALSE;
8825 }
8826
8827 /* Atomic brackets */
8828
8829 else if (op == OP_ONCE || op == OP_ONCE_NC)
8830 {
8831 if (!is_startline(scode, bracket_map, cd, atomcount + 1, inassert)) return FALSE;
8832 }
8833
8834 /* .* means "start at start or after \n" if it isn't in atomic brackets or
8835 brackets that may be referenced or an assertion, as long as the pattern does
8836 not contain *PRUNE or *SKIP, because these break the feature. Consider, for
8837 example, /.*?a(*PRUNE)b/ with the subject "aab", which matches "ab", i.e.
8838 not at the start of a line. */
8839
8840 else if (op == OP_TYPESTAR || op == OP_TYPEMINSTAR || op == OP_TYPEPOSSTAR)
8841 {
8842 if (scode[1] != OP_ANY || (bracket_map & cd->backref_map) != 0 ||
8843 atomcount > 0 || cd->had_pruneorskip || inassert)
8844 return FALSE;
8845 }
8846
8847 /* Check for explicit circumflex; anything else gives a FALSE result. Note
8848 in particular that this includes atomic brackets OP_ONCE and OP_ONCE_NC
8849 because the number of characters matched by .* cannot be adjusted inside
8850 them. */
8851
8852 else if (op != OP_CIRC && op != OP_CIRCM) return FALSE;
8853
8854 /* Move on to the next alternative */
8855
8856 code += GET(code, 1);
8857 }
8858 while (*code == OP_ALT); /* Loop for each alternative */
8859 return TRUE;
8860 }
8861
8862
8863
8864 /*************************************************
8865 * Check for asserted fixed first char *
8866 *************************************************/
8867
8868 /* During compilation, the "first char" settings from forward assertions are
8869 discarded, because they can cause conflicts with actual literals that follow.
8870 However, if we end up without a first char setting for an unanchored pattern,
8871 it is worth scanning the regex to see if there is an initial asserted first
8872 char. If all branches start with the same asserted char, or with a
8873 non-conditional bracket all of whose alternatives start with the same asserted
8874 char (recurse ad lib), then we return that char, with the flags set to zero or
8875 REQ_CASELESS; otherwise return zero with REQ_NONE in the flags.
8876
8877 Arguments:
8878 code points to start of expression (the bracket)
8879 flags points to the first char flags, or to REQ_NONE
8880 inassert TRUE if in an assertion
8881
8882 Returns: the fixed first char, or 0 with REQ_NONE in flags
8883 */
8884
8885 static pcre_uint32
find_firstassertedchar(const pcre_uchar * code,pcre_int32 * flags,BOOL inassert)8886 find_firstassertedchar(const pcre_uchar *code, pcre_int32 *flags,
8887 BOOL inassert)
8888 {
8889 register pcre_uint32 c = 0;
8890 int cflags = REQ_NONE;
8891
8892 *flags = REQ_NONE;
8893 do {
8894 pcre_uint32 d;
8895 int dflags;
8896 int xl = (*code == OP_CBRA || *code == OP_SCBRA ||
8897 *code == OP_CBRAPOS || *code == OP_SCBRAPOS)? IMM2_SIZE:0;
8898 const pcre_uchar *scode = first_significant_code(code + 1+LINK_SIZE + xl,
8899 TRUE);
8900 register pcre_uchar op = *scode;
8901
8902 switch(op)
8903 {
8904 default:
8905 return 0;
8906
8907 case OP_BRA:
8908 case OP_BRAPOS:
8909 case OP_CBRA:
8910 case OP_SCBRA:
8911 case OP_CBRAPOS:
8912 case OP_SCBRAPOS:
8913 case OP_ASSERT:
8914 case OP_ONCE:
8915 case OP_ONCE_NC:
8916 d = find_firstassertedchar(scode, &dflags, op == OP_ASSERT);
8917 if (dflags < 0)
8918 return 0;
8919 if (cflags < 0) { c = d; cflags = dflags; } else if (c != d || cflags != dflags) return 0;
8920 break;
8921
8922 case OP_EXACT:
8923 scode += IMM2_SIZE;
8924 /* Fall through */
8925
8926 case OP_CHAR:
8927 case OP_PLUS:
8928 case OP_MINPLUS:
8929 case OP_POSPLUS:
8930 if (!inassert) return 0;
8931 if (cflags < 0) { c = scode[1]; cflags = 0; }
8932 else if (c != scode[1]) return 0;
8933 break;
8934
8935 case OP_EXACTI:
8936 scode += IMM2_SIZE;
8937 /* Fall through */
8938
8939 case OP_CHARI:
8940 case OP_PLUSI:
8941 case OP_MINPLUSI:
8942 case OP_POSPLUSI:
8943 if (!inassert) return 0;
8944 if (cflags < 0) { c = scode[1]; cflags = REQ_CASELESS; }
8945 else if (c != scode[1]) return 0;
8946 break;
8947 }
8948
8949 code += GET(code, 1);
8950 }
8951 while (*code == OP_ALT);
8952
8953 *flags = cflags;
8954 return c;
8955 }
8956
8957
8958
8959 /*************************************************
8960 * Add an entry to the name/number table *
8961 *************************************************/
8962
8963 /* This function is called between compiling passes to add an entry to the
8964 name/number table, maintaining alphabetical order. Checking for permitted
8965 and forbidden duplicates has already been done.
8966
8967 Arguments:
8968 cd the compile data block
8969 name the name to add
8970 length the length of the name
8971 groupno the group number
8972
8973 Returns: nothing
8974 */
8975
8976 static void
add_name(compile_data * cd,const pcre_uchar * name,int length,unsigned int groupno)8977 add_name(compile_data *cd, const pcre_uchar *name, int length,
8978 unsigned int groupno)
8979 {
8980 int i;
8981 pcre_uchar *slot = cd->name_table;
8982
8983 for (i = 0; i < cd->names_found; i++)
8984 {
8985 int crc = memcmp(name, slot+IMM2_SIZE, IN_UCHARS(length));
8986 if (crc == 0 && slot[IMM2_SIZE+length] != 0)
8987 crc = -1; /* Current name is a substring */
8988
8989 /* Make space in the table and break the loop for an earlier name. For a
8990 duplicate or later name, carry on. We do this for duplicates so that in the
8991 simple case (when ?(| is not used) they are in order of their numbers. In all
8992 cases they are in the order in which they appear in the pattern. */
8993
8994 if (crc < 0)
8995 {
8996 memmove(slot + cd->name_entry_size, slot,
8997 IN_UCHARS((cd->names_found - i) * cd->name_entry_size));
8998 break;
8999 }
9000
9001 /* Continue the loop for a later or duplicate name */
9002
9003 slot += cd->name_entry_size;
9004 }
9005
9006 PUT2(slot, 0, groupno);
9007 memcpy(slot + IMM2_SIZE, name, IN_UCHARS(length));
9008 slot[IMM2_SIZE + length] = 0;
9009 cd->names_found++;
9010 }
9011
9012
9013
9014 /*************************************************
9015 * Compile a Regular Expression *
9016 *************************************************/
9017
9018 /* This function takes a string and returns a pointer to a block of store
9019 holding a compiled version of the expression. The original API for this
9020 function had no error code return variable; it is retained for backwards
9021 compatibility. The new function is given a new name.
9022
9023 Arguments:
9024 pattern the regular expression
9025 options various option bits
9026 errorcodeptr pointer to error code variable (pcre_compile2() only)
9027 can be NULL if you don't want a code value
9028 errorptr pointer to pointer to error text
9029 erroroffset ptr offset in pattern where error was detected
9030 tables pointer to character tables or NULL
9031
9032 Returns: pointer to compiled data block, or NULL on error,
9033 with errorptr and erroroffset set
9034 */
9035
9036 #if defined COMPILE_PCRE8
9037 PCRE_EXP_DEFN pcre * PCRE_CALL_CONVENTION
pcre_compile(const char * pattern,int options,const char ** errorptr,int * erroroffset,const unsigned char * tables)9038 pcre_compile(const char *pattern, int options, const char **errorptr,
9039 int *erroroffset, const unsigned char *tables)
9040 #elif defined COMPILE_PCRE16
9041 PCRE_EXP_DEFN pcre16 * PCRE_CALL_CONVENTION
9042 pcre16_compile(PCRE_SPTR16 pattern, int options, const char **errorptr,
9043 int *erroroffset, const unsigned char *tables)
9044 #elif defined COMPILE_PCRE32
9045 PCRE_EXP_DEFN pcre32 * PCRE_CALL_CONVENTION
9046 pcre32_compile(PCRE_SPTR32 pattern, int options, const char **errorptr,
9047 int *erroroffset, const unsigned char *tables)
9048 #endif
9049 {
9050 #if defined COMPILE_PCRE8
9051 return pcre_compile2(pattern, options, NULL, errorptr, erroroffset, tables);
9052 #elif defined COMPILE_PCRE16
9053 return pcre16_compile2(pattern, options, NULL, errorptr, erroroffset, tables);
9054 #elif defined COMPILE_PCRE32
9055 return pcre32_compile2(pattern, options, NULL, errorptr, erroroffset, tables);
9056 #endif
9057 }
9058
9059
9060 #if defined COMPILE_PCRE8
9061 PCRE_EXP_DEFN pcre * PCRE_CALL_CONVENTION
pcre_compile2(const char * pattern,int options,int * errorcodeptr,const char ** errorptr,int * erroroffset,const unsigned char * tables)9062 pcre_compile2(const char *pattern, int options, int *errorcodeptr,
9063 const char **errorptr, int *erroroffset, const unsigned char *tables)
9064 #elif defined COMPILE_PCRE16
9065 PCRE_EXP_DEFN pcre16 * PCRE_CALL_CONVENTION
9066 pcre16_compile2(PCRE_SPTR16 pattern, int options, int *errorcodeptr,
9067 const char **errorptr, int *erroroffset, const unsigned char *tables)
9068 #elif defined COMPILE_PCRE32
9069 PCRE_EXP_DEFN pcre32 * PCRE_CALL_CONVENTION
9070 pcre32_compile2(PCRE_SPTR32 pattern, int options, int *errorcodeptr,
9071 const char **errorptr, int *erroroffset, const unsigned char *tables)
9072 #endif
9073 {
9074 REAL_PCRE *re;
9075 int length = 1; /* For final END opcode */
9076 pcre_int32 firstcharflags, reqcharflags;
9077 pcre_uint32 firstchar, reqchar;
9078 pcre_uint32 limit_match = PCRE_UINT32_MAX;
9079 pcre_uint32 limit_recursion = PCRE_UINT32_MAX;
9080 int newline;
9081 int errorcode = 0;
9082 int skipatstart = 0;
9083 BOOL utf;
9084 BOOL never_utf = FALSE;
9085 size_t size;
9086 pcre_uchar *code;
9087 const pcre_uchar *codestart;
9088 const pcre_uchar *ptr;
9089 compile_data compile_block;
9090 compile_data *cd = &compile_block;
9091
9092 /* This space is used for "compiling" into during the first phase, when we are
9093 computing the amount of memory that is needed. Compiled items are thrown away
9094 as soon as possible, so that a fairly large buffer should be sufficient for
9095 this purpose. The same space is used in the second phase for remembering where
9096 to fill in forward references to subpatterns. That may overflow, in which case
9097 new memory is obtained from malloc(). */
9098
9099 pcre_uchar cworkspace[COMPILE_WORK_SIZE];
9100
9101 /* This vector is used for remembering name groups during the pre-compile. In a
9102 similar way to cworkspace, it can be expanded using malloc() if necessary. */
9103
9104 named_group named_groups[NAMED_GROUP_LIST_SIZE];
9105 cd->named_groups = named_groups;
9106 cd->named_group_list_size = NAMED_GROUP_LIST_SIZE;
9107
9108 /* Set this early so that early errors get offset 0. */
9109
9110 ptr = (const pcre_uchar *)pattern;
9111
9112 /* We can't pass back an error message if errorptr is NULL; I guess the best we
9113 can do is just return NULL, but we can set a code value if there is a code
9114 pointer. */
9115
9116 if (errorptr == NULL)
9117 {
9118 if (errorcodeptr != NULL) *errorcodeptr = 99;
9119 return NULL;
9120 }
9121
9122 *errorptr = NULL;
9123 if (errorcodeptr != NULL) *errorcodeptr = ERR0;
9124
9125 /* However, we can give a message for this error */
9126
9127 if (erroroffset == NULL)
9128 {
9129 errorcode = ERR16;
9130 goto PCRE_EARLY_ERROR_RETURN2;
9131 }
9132
9133 *erroroffset = 0;
9134
9135 /* Set up pointers to the individual character tables */
9136
9137 if (tables == NULL) tables = PRIV(default_tables);
9138 cd->lcc = tables + lcc_offset;
9139 cd->fcc = tables + fcc_offset;
9140 cd->cbits = tables + cbits_offset;
9141 cd->ctypes = tables + ctypes_offset;
9142
9143 /* Check that all undefined public option bits are zero */
9144
9145 if ((options & ~PUBLIC_COMPILE_OPTIONS) != 0)
9146 {
9147 errorcode = ERR17;
9148 goto PCRE_EARLY_ERROR_RETURN;
9149 }
9150
9151 /* If PCRE_NEVER_UTF is set, remember it. */
9152
9153 if ((options & PCRE_NEVER_UTF) != 0) never_utf = TRUE;
9154
9155 /* Check for global one-time settings at the start of the pattern, and remember
9156 the offset for later. */
9157
9158 cd->external_flags = 0; /* Initialize here for LIMIT_MATCH/RECURSION */
9159
9160 while (ptr[skipatstart] == CHAR_LEFT_PARENTHESIS &&
9161 ptr[skipatstart+1] == CHAR_ASTERISK)
9162 {
9163 int newnl = 0;
9164 int newbsr = 0;
9165
9166 /* For completeness and backward compatibility, (*UTFn) is supported in the
9167 relevant libraries, but (*UTF) is generic and always supported. Note that
9168 PCRE_UTF8 == PCRE_UTF16 == PCRE_UTF32. */
9169
9170 #ifdef COMPILE_PCRE8
9171 if (STRNCMP_UC_C8(ptr+skipatstart+2, STRING_UTF8_RIGHTPAR, 5) == 0)
9172 { skipatstart += 7; options |= PCRE_UTF8; continue; }
9173 #endif
9174 #ifdef COMPILE_PCRE16
9175 if (STRNCMP_UC_C8(ptr+skipatstart+2, STRING_UTF16_RIGHTPAR, 6) == 0)
9176 { skipatstart += 8; options |= PCRE_UTF16; continue; }
9177 #endif
9178 #ifdef COMPILE_PCRE32
9179 if (STRNCMP_UC_C8(ptr+skipatstart+2, STRING_UTF32_RIGHTPAR, 6) == 0)
9180 { skipatstart += 8; options |= PCRE_UTF32; continue; }
9181 #endif
9182
9183 else if (STRNCMP_UC_C8(ptr+skipatstart+2, STRING_UTF_RIGHTPAR, 4) == 0)
9184 { skipatstart += 6; options |= PCRE_UTF8; continue; }
9185 else if (STRNCMP_UC_C8(ptr+skipatstart+2, STRING_UCP_RIGHTPAR, 4) == 0)
9186 { skipatstart += 6; options |= PCRE_UCP; continue; }
9187 else if (STRNCMP_UC_C8(ptr+skipatstart+2, STRING_NO_AUTO_POSSESS_RIGHTPAR, 16) == 0)
9188 { skipatstart += 18; options |= PCRE_NO_AUTO_POSSESS; continue; }
9189 else if (STRNCMP_UC_C8(ptr+skipatstart+2, STRING_NO_START_OPT_RIGHTPAR, 13) == 0)
9190 { skipatstart += 15; options |= PCRE_NO_START_OPTIMIZE; continue; }
9191
9192 else if (STRNCMP_UC_C8(ptr+skipatstart+2, STRING_LIMIT_MATCH_EQ, 12) == 0)
9193 {
9194 pcre_uint32 c = 0;
9195 int p = skipatstart + 14;
9196 while (isdigit(ptr[p]))
9197 {
9198 if (c > PCRE_UINT32_MAX / 10 - 1) break; /* Integer overflow */
9199 c = c*10 + ptr[p++] - CHAR_0;
9200 }
9201 if (ptr[p++] != CHAR_RIGHT_PARENTHESIS) break;
9202 if (c < limit_match)
9203 {
9204 limit_match = c;
9205 cd->external_flags |= PCRE_MLSET;
9206 }
9207 skipatstart = p;
9208 continue;
9209 }
9210
9211 else if (STRNCMP_UC_C8(ptr+skipatstart+2, STRING_LIMIT_RECURSION_EQ, 16) == 0)
9212 {
9213 pcre_uint32 c = 0;
9214 int p = skipatstart + 18;
9215 while (isdigit(ptr[p]))
9216 {
9217 if (c > PCRE_UINT32_MAX / 10 - 1) break; /* Integer overflow check */
9218 c = c*10 + ptr[p++] - CHAR_0;
9219 }
9220 if (ptr[p++] != CHAR_RIGHT_PARENTHESIS) break;
9221 if (c < limit_recursion)
9222 {
9223 limit_recursion = c;
9224 cd->external_flags |= PCRE_RLSET;
9225 }
9226 skipatstart = p;
9227 continue;
9228 }
9229
9230 if (STRNCMP_UC_C8(ptr+skipatstart+2, STRING_CR_RIGHTPAR, 3) == 0)
9231 { skipatstart += 5; newnl = PCRE_NEWLINE_CR; }
9232 else if (STRNCMP_UC_C8(ptr+skipatstart+2, STRING_LF_RIGHTPAR, 3) == 0)
9233 { skipatstart += 5; newnl = PCRE_NEWLINE_LF; }
9234 else if (STRNCMP_UC_C8(ptr+skipatstart+2, STRING_CRLF_RIGHTPAR, 5) == 0)
9235 { skipatstart += 7; newnl = PCRE_NEWLINE_CR + PCRE_NEWLINE_LF; }
9236 else if (STRNCMP_UC_C8(ptr+skipatstart+2, STRING_ANY_RIGHTPAR, 4) == 0)
9237 { skipatstart += 6; newnl = PCRE_NEWLINE_ANY; }
9238 else if (STRNCMP_UC_C8(ptr+skipatstart+2, STRING_ANYCRLF_RIGHTPAR, 8) == 0)
9239 { skipatstart += 10; newnl = PCRE_NEWLINE_ANYCRLF; }
9240
9241 else if (STRNCMP_UC_C8(ptr+skipatstart+2, STRING_BSR_ANYCRLF_RIGHTPAR, 12) == 0)
9242 { skipatstart += 14; newbsr = PCRE_BSR_ANYCRLF; }
9243 else if (STRNCMP_UC_C8(ptr+skipatstart+2, STRING_BSR_UNICODE_RIGHTPAR, 12) == 0)
9244 { skipatstart += 14; newbsr = PCRE_BSR_UNICODE; }
9245
9246 if (newnl != 0)
9247 options = (options & ~PCRE_NEWLINE_BITS) | newnl;
9248 else if (newbsr != 0)
9249 options = (options & ~(PCRE_BSR_ANYCRLF|PCRE_BSR_UNICODE)) | newbsr;
9250 else break;
9251 }
9252
9253 /* PCRE_UTF(16|32) have the same value as PCRE_UTF8. */
9254 utf = (options & PCRE_UTF8) != 0;
9255 if (utf && never_utf)
9256 {
9257 errorcode = ERR78;
9258 goto PCRE_EARLY_ERROR_RETURN2;
9259 }
9260
9261 /* Can't support UTF unless PCRE has been compiled to include the code. The
9262 return of an error code from PRIV(valid_utf)() is a new feature, introduced in
9263 release 8.13. It is passed back from pcre_[dfa_]exec(), but at the moment is
9264 not used here. */
9265
9266 #ifdef SUPPORT_UTF
9267 if (utf && (options & PCRE_NO_UTF8_CHECK) == 0 &&
9268 (errorcode = PRIV(valid_utf)((PCRE_PUCHAR)pattern, -1, erroroffset)) != 0)
9269 {
9270 #if defined COMPILE_PCRE8
9271 errorcode = ERR44;
9272 #elif defined COMPILE_PCRE16
9273 errorcode = ERR74;
9274 #elif defined COMPILE_PCRE32
9275 errorcode = ERR77;
9276 #endif
9277 goto PCRE_EARLY_ERROR_RETURN2;
9278 }
9279 #else
9280 if (utf)
9281 {
9282 errorcode = ERR32;
9283 goto PCRE_EARLY_ERROR_RETURN;
9284 }
9285 #endif
9286
9287 /* Can't support UCP unless PCRE has been compiled to include the code. */
9288
9289 #ifndef SUPPORT_UCP
9290 if ((options & PCRE_UCP) != 0)
9291 {
9292 errorcode = ERR67;
9293 goto PCRE_EARLY_ERROR_RETURN;
9294 }
9295 #endif
9296
9297 /* Check validity of \R options. */
9298
9299 if ((options & (PCRE_BSR_ANYCRLF|PCRE_BSR_UNICODE)) ==
9300 (PCRE_BSR_ANYCRLF|PCRE_BSR_UNICODE))
9301 {
9302 errorcode = ERR56;
9303 goto PCRE_EARLY_ERROR_RETURN;
9304 }
9305
9306 /* Handle different types of newline. The three bits give seven cases. The
9307 current code allows for fixed one- or two-byte sequences, plus "any" and
9308 "anycrlf". */
9309
9310 switch (options & PCRE_NEWLINE_BITS)
9311 {
9312 case 0: newline = NEWLINE; break; /* Build-time default */
9313 case PCRE_NEWLINE_CR: newline = CHAR_CR; break;
9314 case PCRE_NEWLINE_LF: newline = CHAR_NL; break;
9315 case PCRE_NEWLINE_CR+
9316 PCRE_NEWLINE_LF: newline = (CHAR_CR << 8) | CHAR_NL; break;
9317 case PCRE_NEWLINE_ANY: newline = -1; break;
9318 case PCRE_NEWLINE_ANYCRLF: newline = -2; break;
9319 default: errorcode = ERR56; goto PCRE_EARLY_ERROR_RETURN;
9320 }
9321
9322 if (newline == -2)
9323 {
9324 cd->nltype = NLTYPE_ANYCRLF;
9325 }
9326 else if (newline < 0)
9327 {
9328 cd->nltype = NLTYPE_ANY;
9329 }
9330 else
9331 {
9332 cd->nltype = NLTYPE_FIXED;
9333 if (newline > 255)
9334 {
9335 cd->nllen = 2;
9336 cd->nl[0] = (newline >> 8) & 255;
9337 cd->nl[1] = newline & 255;
9338 }
9339 else
9340 {
9341 cd->nllen = 1;
9342 cd->nl[0] = newline;
9343 }
9344 }
9345
9346 /* Maximum back reference and backref bitmap. The bitmap records up to 31 back
9347 references to help in deciding whether (.*) can be treated as anchored or not.
9348 */
9349
9350 cd->top_backref = 0;
9351 cd->backref_map = 0;
9352
9353 /* Reflect pattern for debugging output */
9354
9355 DPRINTF(("------------------------------------------------------------------\n"));
9356 #ifdef PCRE_DEBUG
9357 print_puchar(stdout, (PCRE_PUCHAR)pattern);
9358 #endif
9359 DPRINTF(("\n"));
9360
9361 /* Pretend to compile the pattern while actually just accumulating the length
9362 of memory required. This behaviour is triggered by passing a non-NULL final
9363 argument to compile_regex(). We pass a block of workspace (cworkspace) for it
9364 to compile parts of the pattern into; the compiled code is discarded when it is
9365 no longer needed, so hopefully this workspace will never overflow, though there
9366 is a test for its doing so. */
9367
9368 cd->bracount = cd->final_bracount = 0;
9369 cd->names_found = 0;
9370 cd->name_entry_size = 0;
9371 cd->name_table = NULL;
9372 cd->dupnames = FALSE;
9373 cd->dupgroups = FALSE;
9374 cd->namedrefcount = 0;
9375 cd->start_code = cworkspace;
9376 cd->hwm = cworkspace;
9377 cd->iscondassert = FALSE;
9378 cd->start_workspace = cworkspace;
9379 cd->workspace_size = COMPILE_WORK_SIZE;
9380 cd->start_pattern = (const pcre_uchar *)pattern;
9381 cd->end_pattern = (const pcre_uchar *)(pattern + STRLEN_UC((const pcre_uchar *)pattern));
9382 cd->req_varyopt = 0;
9383 cd->parens_depth = 0;
9384 cd->assert_depth = 0;
9385 cd->max_lookbehind = 0;
9386 cd->external_options = options;
9387 cd->open_caps = NULL;
9388
9389 /* Now do the pre-compile. On error, errorcode will be set non-zero, so we
9390 don't need to look at the result of the function here. The initial options have
9391 been put into the cd block so that they can be changed if an option setting is
9392 found within the regex right at the beginning. Bringing initial option settings
9393 outside can help speed up starting point checks. */
9394
9395 ptr += skipatstart;
9396 code = cworkspace;
9397 *code = OP_BRA;
9398
9399 (void)compile_regex(cd->external_options, &code, &ptr, &errorcode, FALSE,
9400 FALSE, 0, 0, &firstchar, &firstcharflags, &reqchar, &reqcharflags, NULL,
9401 cd, &length);
9402 if (errorcode != 0) goto PCRE_EARLY_ERROR_RETURN;
9403
9404 DPRINTF(("end pre-compile: length=%d workspace=%d\n", length,
9405 (int)(cd->hwm - cworkspace)));
9406
9407 if (length > MAX_PATTERN_SIZE)
9408 {
9409 errorcode = ERR20;
9410 goto PCRE_EARLY_ERROR_RETURN;
9411 }
9412
9413 /* Compute the size of the data block for storing the compiled pattern. Integer
9414 overflow should no longer be possible because nowadays we limit the maximum
9415 value of cd->names_found and cd->name_entry_size. */
9416
9417 size = sizeof(REAL_PCRE) +
9418 (length + cd->names_found * cd->name_entry_size) * sizeof(pcre_uchar);
9419
9420 /* Get the memory. */
9421
9422 re = (REAL_PCRE *)(PUBL(malloc))(size);
9423 if (re == NULL)
9424 {
9425 errorcode = ERR21;
9426 goto PCRE_EARLY_ERROR_RETURN;
9427 }
9428
9429 /* Put in the magic number, and save the sizes, initial options, internal
9430 flags, and character table pointer. NULL is used for the default character
9431 tables. The nullpad field is at the end; it's there to help in the case when a
9432 regex compiled on a system with 4-byte pointers is run on another with 8-byte
9433 pointers. */
9434
9435 re->magic_number = MAGIC_NUMBER;
9436 re->size = (int)size;
9437 re->options = cd->external_options;
9438 re->flags = cd->external_flags;
9439 re->limit_match = limit_match;
9440 re->limit_recursion = limit_recursion;
9441 re->first_char = 0;
9442 re->req_char = 0;
9443 re->name_table_offset = sizeof(REAL_PCRE) / sizeof(pcre_uchar);
9444 re->name_entry_size = cd->name_entry_size;
9445 re->name_count = cd->names_found;
9446 re->ref_count = 0;
9447 re->tables = (tables == PRIV(default_tables))? NULL : tables;
9448 re->nullpad = NULL;
9449 #ifdef COMPILE_PCRE32
9450 re->dummy = 0;
9451 #else
9452 re->dummy1 = re->dummy2 = re->dummy3 = 0;
9453 #endif
9454
9455 /* The starting points of the name/number translation table and of the code are
9456 passed around in the compile data block. The start/end pattern and initial
9457 options are already set from the pre-compile phase, as is the name_entry_size
9458 field. Reset the bracket count and the names_found field. Also reset the hwm
9459 field; this time it's used for remembering forward references to subpatterns.
9460 */
9461
9462 cd->final_bracount = cd->bracount; /* Save for checking forward references */
9463 cd->parens_depth = 0;
9464 cd->assert_depth = 0;
9465 cd->bracount = 0;
9466 cd->max_lookbehind = 0;
9467 cd->name_table = (pcre_uchar *)re + re->name_table_offset;
9468 codestart = cd->name_table + re->name_entry_size * re->name_count;
9469 cd->start_code = codestart;
9470 cd->hwm = (pcre_uchar *)(cd->start_workspace);
9471 cd->iscondassert = FALSE;
9472 cd->req_varyopt = 0;
9473 cd->had_accept = FALSE;
9474 cd->had_pruneorskip = FALSE;
9475 cd->check_lookbehind = FALSE;
9476 cd->open_caps = NULL;
9477
9478 /* If any named groups were found, create the name/number table from the list
9479 created in the first pass. */
9480
9481 if (cd->names_found > 0)
9482 {
9483 int i = cd->names_found;
9484 named_group *ng = cd->named_groups;
9485 cd->names_found = 0;
9486 for (; i > 0; i--, ng++)
9487 add_name(cd, ng->name, ng->length, ng->number);
9488 if (cd->named_group_list_size > NAMED_GROUP_LIST_SIZE)
9489 (PUBL(free))((void *)cd->named_groups);
9490 cd->named_group_list_size = 0; /* So we don't free it twice */
9491 }
9492
9493 /* Set up a starting, non-extracting bracket, then compile the expression. On
9494 error, errorcode will be set non-zero, so we don't need to look at the result
9495 of the function here. */
9496
9497 ptr = (const pcre_uchar *)pattern + skipatstart;
9498 code = (pcre_uchar *)codestart;
9499 *code = OP_BRA;
9500 (void)compile_regex(re->options, &code, &ptr, &errorcode, FALSE, FALSE, 0, 0,
9501 &firstchar, &firstcharflags, &reqchar, &reqcharflags, NULL, cd, NULL);
9502 re->top_bracket = cd->bracount;
9503 re->top_backref = cd->top_backref;
9504 re->max_lookbehind = cd->max_lookbehind;
9505 re->flags = cd->external_flags | PCRE_MODE;
9506
9507 if (cd->had_accept)
9508 {
9509 reqchar = 0; /* Must disable after (*ACCEPT) */
9510 reqcharflags = REQ_NONE;
9511 }
9512
9513 /* If not reached end of pattern on success, there's an excess bracket. */
9514
9515 if (errorcode == 0 && *ptr != CHAR_NULL) errorcode = ERR22;
9516
9517 /* Fill in the terminating state and check for disastrous overflow, but
9518 if debugging, leave the test till after things are printed out. */
9519
9520 *code++ = OP_END;
9521
9522 #ifndef PCRE_DEBUG
9523 if (code - codestart > length) errorcode = ERR23;
9524 #endif
9525
9526 #ifdef SUPPORT_VALGRIND
9527 /* If the estimated length exceeds the really used length, mark the extra
9528 allocated memory as unaddressable, so that any out-of-bound reads can be
9529 detected. */
9530 VALGRIND_MAKE_MEM_NOACCESS(code, (length - (code - codestart)) * sizeof(pcre_uchar));
9531 #endif
9532
9533 /* Fill in any forward references that are required. There may be repeated
9534 references; optimize for them, as searching a large regex takes time. */
9535
9536 if (cd->hwm > cd->start_workspace)
9537 {
9538 int prev_recno = -1;
9539 const pcre_uchar *groupptr = NULL;
9540 while (errorcode == 0 && cd->hwm > cd->start_workspace)
9541 {
9542 int offset, recno;
9543 cd->hwm -= LINK_SIZE;
9544 offset = GET(cd->hwm, 0);
9545
9546 /* Check that the hwm handling hasn't gone wrong. This whole area is
9547 rewritten in PCRE2 because there are some obscure cases. */
9548
9549 if (offset == 0 || codestart[offset-1] != OP_RECURSE)
9550 {
9551 errorcode = ERR10;
9552 break;
9553 }
9554
9555 recno = GET(codestart, offset);
9556 if (recno != prev_recno)
9557 {
9558 groupptr = PRIV(find_bracket)(codestart, utf, recno);
9559 prev_recno = recno;
9560 }
9561 if (groupptr == NULL) errorcode = ERR53;
9562 else PUT(((pcre_uchar *)codestart), offset, (int)(groupptr - codestart));
9563 }
9564 }
9565
9566 /* If the workspace had to be expanded, free the new memory. Set the pointer to
9567 NULL to indicate that forward references have been filled in. */
9568
9569 if (cd->workspace_size > COMPILE_WORK_SIZE)
9570 (PUBL(free))((void *)cd->start_workspace);
9571 cd->start_workspace = NULL;
9572
9573 /* Give an error if there's back reference to a non-existent capturing
9574 subpattern. */
9575
9576 if (errorcode == 0 && re->top_backref > re->top_bracket) errorcode = ERR15;
9577
9578 /* Unless disabled, check whether any single character iterators can be
9579 auto-possessified. The function overwrites the appropriate opcode values, so
9580 the type of the pointer must be cast. NOTE: the intermediate variable "temp" is
9581 used in this code because at least one compiler gives a warning about loss of
9582 "const" attribute if the cast (pcre_uchar *)codestart is used directly in the
9583 function call. */
9584
9585 if (errorcode == 0 && (options & PCRE_NO_AUTO_POSSESS) == 0)
9586 {
9587 pcre_uchar *temp = (pcre_uchar *)codestart;
9588 auto_possessify(temp, utf, cd);
9589 }
9590
9591 /* If there were any lookbehind assertions that contained OP_RECURSE
9592 (recursions or subroutine calls), a flag is set for them to be checked here,
9593 because they may contain forward references. Actual recursions cannot be fixed
9594 length, but subroutine calls can. It is done like this so that those without
9595 OP_RECURSE that are not fixed length get a diagnosic with a useful offset. The
9596 exceptional ones forgo this. We scan the pattern to check that they are fixed
9597 length, and set their lengths. */
9598
9599 if (errorcode == 0 && cd->check_lookbehind)
9600 {
9601 pcre_uchar *cc = (pcre_uchar *)codestart;
9602
9603 /* Loop, searching for OP_REVERSE items, and process those that do not have
9604 their length set. (Actually, it will also re-process any that have a length
9605 of zero, but that is a pathological case, and it does no harm.) When we find
9606 one, we temporarily terminate the branch it is in while we scan it. */
9607
9608 for (cc = (pcre_uchar *)PRIV(find_bracket)(codestart, utf, -1);
9609 cc != NULL;
9610 cc = (pcre_uchar *)PRIV(find_bracket)(cc, utf, -1))
9611 {
9612 if (GET(cc, 1) == 0)
9613 {
9614 int fixed_length;
9615 pcre_uchar *be = cc - 1 - LINK_SIZE + GET(cc, -LINK_SIZE);
9616 int end_op = *be;
9617 *be = OP_END;
9618 fixed_length = find_fixedlength(cc, (re->options & PCRE_UTF8) != 0, TRUE,
9619 cd, NULL);
9620 *be = end_op;
9621 DPRINTF(("fixed length = %d\n", fixed_length));
9622 if (fixed_length < 0)
9623 {
9624 errorcode = (fixed_length == -2)? ERR36 :
9625 (fixed_length == -4)? ERR70 : ERR25;
9626 break;
9627 }
9628 if (fixed_length > cd->max_lookbehind) cd->max_lookbehind = fixed_length;
9629 PUT(cc, 1, fixed_length);
9630 }
9631 cc += 1 + LINK_SIZE;
9632 }
9633 }
9634
9635 /* Failed to compile, or error while post-processing */
9636
9637 if (errorcode != 0)
9638 {
9639 (PUBL(free))(re);
9640 PCRE_EARLY_ERROR_RETURN:
9641 if (cd->named_group_list_size > NAMED_GROUP_LIST_SIZE)
9642 (PUBL(free))((void *)cd->named_groups);
9643 *erroroffset = (int)(ptr - (const pcre_uchar *)pattern);
9644 PCRE_EARLY_ERROR_RETURN2:
9645 *errorptr = find_error_text(errorcode);
9646 if (errorcodeptr != NULL) *errorcodeptr = errorcode;
9647 return NULL;
9648 }
9649
9650 /* If the anchored option was not passed, set the flag if we can determine that
9651 the pattern is anchored by virtue of ^ characters or \A or anything else, such
9652 as starting with non-atomic .* when DOTALL is set and there are no occurrences
9653 of *PRUNE or *SKIP.
9654
9655 Otherwise, if we know what the first byte has to be, save it, because that
9656 speeds up unanchored matches no end. If not, see if we can set the
9657 PCRE_STARTLINE flag. This is helpful for multiline matches when all branches
9658 start with ^. and also when all branches start with non-atomic .* for
9659 non-DOTALL matches when *PRUNE and SKIP are not present. */
9660
9661 if ((re->options & PCRE_ANCHORED) == 0)
9662 {
9663 if (is_anchored(codestart, 0, cd, 0)) re->options |= PCRE_ANCHORED;
9664 else
9665 {
9666 if (firstcharflags < 0)
9667 firstchar = find_firstassertedchar(codestart, &firstcharflags, FALSE);
9668 if (firstcharflags >= 0) /* Remove caseless flag for non-caseable chars */
9669 {
9670 #if defined COMPILE_PCRE8
9671 re->first_char = firstchar & 0xff;
9672 #elif defined COMPILE_PCRE16
9673 re->first_char = firstchar & 0xffff;
9674 #elif defined COMPILE_PCRE32
9675 re->first_char = firstchar;
9676 #endif
9677 if ((firstcharflags & REQ_CASELESS) != 0)
9678 {
9679 #if defined SUPPORT_UCP && !(defined COMPILE_PCRE8)
9680 /* We ignore non-ASCII first chars in 8 bit mode. */
9681 if (utf)
9682 {
9683 if (re->first_char < 128)
9684 {
9685 if (cd->fcc[re->first_char] != re->first_char)
9686 re->flags |= PCRE_FCH_CASELESS;
9687 }
9688 else if (UCD_OTHERCASE(re->first_char) != re->first_char)
9689 re->flags |= PCRE_FCH_CASELESS;
9690 }
9691 else
9692 #endif
9693 if (MAX_255(re->first_char)
9694 && cd->fcc[re->first_char] != re->first_char)
9695 re->flags |= PCRE_FCH_CASELESS;
9696 }
9697
9698 re->flags |= PCRE_FIRSTSET;
9699 }
9700
9701 else if (is_startline(codestart, 0, cd, 0, FALSE)) re->flags |= PCRE_STARTLINE;
9702 }
9703 }
9704
9705 /* For an anchored pattern, we use the "required byte" only if it follows a
9706 variable length item in the regex. Remove the caseless flag for non-caseable
9707 bytes. */
9708
9709 if (reqcharflags >= 0 &&
9710 ((re->options & PCRE_ANCHORED) == 0 || (reqcharflags & REQ_VARY) != 0))
9711 {
9712 #if defined COMPILE_PCRE8
9713 re->req_char = reqchar & 0xff;
9714 #elif defined COMPILE_PCRE16
9715 re->req_char = reqchar & 0xffff;
9716 #elif defined COMPILE_PCRE32
9717 re->req_char = reqchar;
9718 #endif
9719 if ((reqcharflags & REQ_CASELESS) != 0)
9720 {
9721 #if defined SUPPORT_UCP && !(defined COMPILE_PCRE8)
9722 /* We ignore non-ASCII first chars in 8 bit mode. */
9723 if (utf)
9724 {
9725 if (re->req_char < 128)
9726 {
9727 if (cd->fcc[re->req_char] != re->req_char)
9728 re->flags |= PCRE_RCH_CASELESS;
9729 }
9730 else if (UCD_OTHERCASE(re->req_char) != re->req_char)
9731 re->flags |= PCRE_RCH_CASELESS;
9732 }
9733 else
9734 #endif
9735 if (MAX_255(re->req_char) && cd->fcc[re->req_char] != re->req_char)
9736 re->flags |= PCRE_RCH_CASELESS;
9737 }
9738
9739 re->flags |= PCRE_REQCHSET;
9740 }
9741
9742 /* Print out the compiled data if debugging is enabled. This is never the
9743 case when building a production library. */
9744
9745 #ifdef PCRE_DEBUG
9746 printf("Length = %d top_bracket = %d top_backref = %d\n",
9747 length, re->top_bracket, re->top_backref);
9748
9749 printf("Options=%08x\n", re->options);
9750
9751 if ((re->flags & PCRE_FIRSTSET) != 0)
9752 {
9753 pcre_uchar ch = re->first_char;
9754 const char *caseless =
9755 ((re->flags & PCRE_FCH_CASELESS) == 0)? "" : " (caseless)";
9756 if (PRINTABLE(ch)) printf("First char = %c%s\n", ch, caseless);
9757 else printf("First char = \\x%02x%s\n", ch, caseless);
9758 }
9759
9760 if ((re->flags & PCRE_REQCHSET) != 0)
9761 {
9762 pcre_uchar ch = re->req_char;
9763 const char *caseless =
9764 ((re->flags & PCRE_RCH_CASELESS) == 0)? "" : " (caseless)";
9765 if (PRINTABLE(ch)) printf("Req char = %c%s\n", ch, caseless);
9766 else printf("Req char = \\x%02x%s\n", ch, caseless);
9767 }
9768
9769 #if defined COMPILE_PCRE8
9770 pcre_printint((pcre *)re, stdout, TRUE);
9771 #elif defined COMPILE_PCRE16
9772 pcre16_printint((pcre *)re, stdout, TRUE);
9773 #elif defined COMPILE_PCRE32
9774 pcre32_printint((pcre *)re, stdout, TRUE);
9775 #endif
9776
9777 /* This check is done here in the debugging case so that the code that
9778 was compiled can be seen. */
9779
9780 if (code - codestart > length)
9781 {
9782 (PUBL(free))(re);
9783 *errorptr = find_error_text(ERR23);
9784 *erroroffset = ptr - (pcre_uchar *)pattern;
9785 if (errorcodeptr != NULL) *errorcodeptr = ERR23;
9786 return NULL;
9787 }
9788 #endif /* PCRE_DEBUG */
9789
9790 /* Check for a pattern than can match an empty string, so that this information
9791 can be provided to applications. */
9792
9793 do
9794 {
9795 if (could_be_empty_branch(codestart, code, utf, cd, NULL))
9796 {
9797 re->flags |= PCRE_MATCH_EMPTY;
9798 break;
9799 }
9800 codestart += GET(codestart, 1);
9801 }
9802 while (*codestart == OP_ALT);
9803
9804 #if defined COMPILE_PCRE8
9805 return (pcre *)re;
9806 #elif defined COMPILE_PCRE16
9807 return (pcre16 *)re;
9808 #elif defined COMPILE_PCRE32
9809 return (pcre32 *)re;
9810 #endif
9811 }
9812
9813 /* End of pcre_compile.c */
9814